An interview with Sven Treitel

Coordinated by: Satinder Chopra | Photos courtesy: Penny Colton

Sven Treitel, a well-known name in the geophysical world, delivered the keynote address at this year’s inaugural CSEG Doodle Train week (November 4th, 2002, Calgary). Recipient of numerous learned society awards, Sven has made distinct contributions to exploration geophysics, especially in digital seismic data processing. His talk brought back memories on how the process of deconvolution evolved and was implemented in the 1960s and 1970s, and how it became an important tool in seismic data processing in the 1980s and beyond.

Before his talk, when approached for this interview, Sven graciously made time that evening, in spite of his busy schedule. Larry Lines and Sam Gray, Sven’s old associates and friends, also joined in the conversation. Sven is soft-spoken, but was very responsive and encouraging in answering our questions. The following are excerpts from the conversation.

[Satinder]: Sven, could you tell us about your educational background and experience?

[Sven]: I grew up in South America, in Argentina. When I finished high school, I decided to come to the US because of the political situation there. I was accepted by MIT and Caltech. Caltech seemed to be very far away and MIT was on the Atlantic coast and closer. That is why I chose MIT. I stayed there for 9 years and ended up with a Ph.D. in Geophysics in 1958. I decided to go into industrial work because of the academic politics of which I saw a lot as a graduate student. It did not appeal to me and I thought I would be better off in the industry. And this was really the case until the oil industry entered its most difficult stage about 10 years ago. I began to ask myself if I would have done better to go into the academic world. It was a bit late, but I did go back to academia. I spent two semesters at MIT and one whole year at Karlsruhe University in Germany.

[Satinder]: What are you doing now, in terms of assignments?

[Sven]: I guess I make assignments for myself. I am officially retired, but I do a lot of consulting, primarily for contractors. In addition to consulting, I like to give short courses and lectures on subjects that I know something about. The advantage there is that a lot of subjects I like to talk about are no longer in the mainstream. They should be, but they are not. This morning’s talk is a good example – while people use deconvolution, they have forgotten where it came from and they have forgotten its limitations. In fact, if I think of my career in the past, I have always liked to work on subjects where most people were not working, because if you work in areas where every Tom, Dick and Harry is working, it’s very crowded, but if you are working in some other area where not that many people are active, then you are better off because it is easier to become known and to have an impact. I won’t claim that I always did that, but very often that was the case. When I got into, for example, digital processing, very few people thought that was possible. In fact, we had people in this industry who thought that the only kind of filter that would work is an analog one, made from resistors, capacitors and inductors.

[Satinder]: Looking back at your illustrious career now, if you were to start all over again, would you have done anything differently than what you did?

[Sven]: Well, I think probably I would have gone more into academics, but then who knows upfront. You do not know the future, so you can’t make the right decisions. The reason is that I find it very disappointing nowadays that so many people I worked with are either unemployed or have left the field; or, if they are still around, they are not working to make full use of their knowledge. I consider it a tremendous waste. I find this terribly discouraging. So, if I had to do it all over again, I might not go into an industrial career, but now that I have it behind me I can’t reverse time.

[Sam]: But if you had gone into an academic career, you would just be training people to be in the exact state you just described.

[Sven]: Yes.

[Sam]: That’s what you did at Amoco.

[Sven]: Yes, that’s assuming that I stayed in exploration geophysics. Who knows, I might have ended up in crustal or global geophysics? It is a difficult question to answer.

[Satinder]: I just wanted to find out if you had any regrets.

[Sam]: No, I don’t have any regrets. I had a great career in the industry and I met a lot of people whom I respect. It was a wonderful experience. This morning I met Sam and many others who I’d worked with over the years. It’s not just the science, it’s also the friendships. We make friendships because we share a common craze for geophysics, which I still have, even though I should have lost it.

[Satinder]: Well, it’s been good for our industry too! If you were to squeeze the essence of your experience, how would put it in a few words?

[Sven]: Well, I think my main contribution to this field has been to build a bridge in from signal processing theory on the one hand and to application and exploration geophysics on the other. I did not develop any of the basic signal processing technologies that found their way into geophysics. Where I think I did help, along with people like Enders Robinson and many others, was to build a bridge between the two disciplines, something that was lacking. There is a lot of opportunity for others, younger than me, to do the same nowadays in other evolving fields.

Photo 01
Sam Gray handing over the memento to Sven Treitel.

[Sam]: I think another thing was to show there were more bridges to be built in the industry and academics, for example exploration.

[Sven]: Yes, in fact while walking back from my talk and lunch with Brian Russell, he told me he was doing his Ph.D. thesis in neural networks. Again, there is a large neural net community within the signal processing community, which is not well coupled with exploration geophysics. There are a lot of potential applications for neural network technology in exploration geophysics. But then it takes people who have a foot in both camps to build more bridges. Now some of these efforts will fail. I can tell you plenty of my efforts failed. But some succeed. The point is if you don’t try anything, you don’t fail, but you may not achieve anything either.

[Satinder]: Today we talk a lot about reservoir characterization. Back in the 60s and 70s, how much effort was directed towards this? Did people talk about doing this type of work?

[Sven]: Very little reservoir characterization was done by geologists and geophysicists. Whatever work was done, was done by petroleum engineers. When I began my career in the petroleum industry, communication between petroleum engineers and geophysicists was practically non-existent, even though both were housed in the same building. Now this was inexcusable, but nevertheless it happened. It is only in recent years that reservoir characterization has come into its own. Now, it is a field with a lot of future, particularly the integration of geophysics into reservoir characterization. There was a lot of resistance when this thing began, not just by petroleum engineers, as we would like to believe, but a lot of geophysicists thought they had nothing to contribute. Of course, this was certainly not true.

[Satinder]: You had mentioned in your talk that Enders Robinson accidentally stumbled upon the idea of deconvolution, which he called ‘predictive decomposition’ which was based on a 1938 theorem by the Swedish statistician Herman Wold. This was the so-called “Predictive Decomposition”theorem. If Robinson had not done so, do you think we would be doing deconvolution today?

[Sven]: Well, it is the old question that if Newton hadn’t been around, would somebody else have discovered his laws? My guess is sooner or later somebody would have seen the connection, but it might have been 10 years later. Given the fact that most scientists who have ever lived, have been alive in the past generation or two, I think the probability that somebody else would have discovered this connection is very high. But it happens, when it happens!

[Sam]: It seems a stretch to me to take Wiener’s application of least squares filtering for anti-aircraft purposes, automatic naval gunnery control and turn it completely on its side and apply it to seismic. That seems like a big leap to me.

[Sven]: That is to Robinson’s credit that he saw the connection between typical seismic exploration on the one hand and Wold’s predictive decomposition theorem on the other, which established the theoretical basis of what today we call predictive deconvolution. But one still requires the engine to drive this entire deconvolution procedure. That engine was given by Norbert Wiener’s work and to integrate these ideas was to his credit.

I came into the picture afterwards. I was still an undergraduate when I joined the project. At MIT it was called the Geophysical Analysis Group. The basic ideas were already understood and we were working on their implementation. The material I showed at today’s talk came from that stage of the project. But it is also interesting, as I pointed out in my talk, that the initial successes were followed by failures. In fact, there were oil companies that after 2- 3 years following the initial success cut off grants and the project died and essentially became inactive until the mid 60s when GSI, which is now defunct, resurrected it. GSI had Milo Backus and others on the staff there. He helped revive this technique within the contracting industry, while Robinson and I did the same within a major oil company. It then soon spread to other oil companies. We were by no means the only such resurrectors. There were several other independent groups, for example Manus Foster at Mobil and John Sherwood, then at Chevron, who were also working on these topics.

[Sam]: That was the beginning of the digital age.

[Sven]: Yes, the digital revolution. But there is a funny story. When GSI came out with the first decon filters, the filters were all of length 55, and the people at our lab thought GSI had come up with a magic number for filter length; not 56 mind you, not 53, but they had to be 55 coefficients long! I made attempts to convince the managers that that was an unlikely explanation, because I could not think of any theoretical reason why the filter should contain 55 coefficients. Well, it turned out the architecture of these first digital computers could handle no more then 55 coefficients. So, that was the reason for the number 55!

[Satinder]: Sven, in your talk you took us down memory lane and showed us the results of early decon applications. In the early 70s interpretation was directed basically towards finding structures and we started using deconvolution to help us meet these goals. The awareness about stratigraphic interpretation caught up in the late 70s and the AAPG Memoir 26 published in 1976 helped propagate this. If we had known about stratigraphic interpretation, do you think deconvolution would have helped in the interpretation of stratigraphic anomalies?

[Sven]: Yes, I think so. I view these two techniques to be complementary, one to the other. The decon models I showed this morning were extremely simple. It is just the assumption that the seismic trace can be thought of in terms of a predictable part, which can be removed, and the unpredictable part which one may interpret as the reflectivity function we seek. It is as simple as that. All other geologic consequences, which are many, can be inferred.

I think that, had we known about the stratigraphic effects you mention, the evolution of deconvolution techniques might have been quite different.

[Satinder]: We go through a lot of sophisticated processing these days, such as surface consistent deconvolution, Q-compensation, and so on.

[Sven]: We knew about Q compensation, we knew the Earth did absorb inelastically but we did not know how to correct it properly. That development came much later, say by the late 60s. The theory had been developed.

[Sam]: Still, your model for predictive deconvolution essentially was stratigraphic. So it really is complementary. I mean it aided stratigraphic interpretation.

[Sven]: Well, the question, what would have happened to the evolution of predictive deconvolution had the understanding of stratigraphic interpretation preceded it, my guess, for what it is worth, is that it would have evolved anyway. The basic theorem was stated by Wold, as I mentioned earlier, in 1938. At that time, Wold knew nothing about seismology. In fact, Wold was himself a student of the Swedish statistician Harald Cramer. After Robinson finished his thesis, Wold invited him to spend some time at Uppsala University, where both collaborated for a while.

[Sam]: Sven, everyone knows that you were highly influential in the development of predictive decon. But you were also influential in wave theory, inverse theory in general and you were around and influential in a number of other topics. Could you tell us a bit about that?

[Sven]: Yes, that is true. There has been more in my career than decon! For example, in the early 70s I saw papers in the Bulletin of the Seismological Society of America, where finite difference techniques were being used to model wave propagation in structurally complex media. I felt these techniques should be injected into exploration geophysics. So, I wrote a very simple code to convince myself that finite difference methods work. But, I also realized that I couldn’t possibly do their development by myself and still work on some of the other projects I promised the management I would work on. So, I told them that they need to hire more people. The first guy hired was a gentleman by the name of Ken Kelly, who really made a name for himself as a finite difference forward modeler. But then we soon recognized that you could use the method of finite differences to image the data as well. Meanwhile, there was another colleague at our lab, Dan Whitmore, who established a reputation in the imaging field; and he is now with ConocoPhillips. He actually came up with a technique to run wave equation backwards, now called reverse time migration. While I was not personally involved in this work, I guess I can take the some credit for bringing these such bright young (at the time!) people into geophysics then. Now of course they have grey hair, and are into geophysics. By the way, neither Kelly nor Whitmore were educated as geophysicists – Kelly is an electrical engineer and Whitmore is a mathematician.

[Sam]: How about Alford?

[Sven]: Oh yes, Alford of Alford rotation fame. Again, I had nothing directly to do with his work, but he was another original finite difference guy. I met him as an undergraduate, recognized him as a very bright young kid and convinced my boss to hire him part time to do programming. But I soon realized he could do far more than write programs. He then got involved with Kelly and myself in finite difference modeling.

[Sven]: But time elapsed, as Sam points out, and I spent more of my time actually helping to build the group at Amoco rather than being directly involved in the research work. I brought in a lot of better people….. A large number of excellent people were recruited.

Photo 02
Doug Uffen (CSEG President), Bill Nickerson (CSEG Doodletrain Chairman), John Bregar (Anadarko), Sven Treitel (TriDekon), Sam Gray(Veritas), Gary Maclean (Anadarko), Larry Lines (Department head, U of Calgary Geology and Geophysics)

[Sam]: As well as me!

[Sven]: … including in particular Leon Thomsen, who was here a while back, and who you also interviewed for this column. I think I can claim at least some of the credit for creating a good environment at the old Amoco lab in those days. We attracted many good scientists to come there to work. Those good times lasted until the mid 80s, when the industry went into a slide, a slide which is continuing today.

[Satinder]: In your talk you mentioned that signal processing research within the oil industry is practically dead.

[Sven]: It is dead right now, but it is resurrectable.

[Satinder]: Everybody says that. Ever since Amoco’s research centre was dismantled, people know that that is the kind of research that needs to be done. But I do not see any sign or an effort in this direction. Even though prominent people like you are telling the industry that we are at a dead end and there is an urgent need to do something, I don’t see anything being done.

[Sven]: Well, it is very gracious of you to say this. In my talk today, I was preaching to the converted; there were not many people in the audience who would disagree that more research in signal processing is necessary. But the people who decided to dismantle the research in this industry were not the ones sitting in the audience. These are the managers with the mentality of the nickel and dime accountants, who now run the large oil companies. Such people have no vision except beyond the next quarterly report. So, this is not going to change unless a new, technically more sophisticated and far-seeing generation of managers takes over this industry. I do not see any indication of this happening soon, but it may happen anyway. It is very hard to predict the future, that is one thing I learnt in the past.

[Satinder]: It is unfortunate.

[Sam]: I think we need to view the present condition as the initial condition and move forward. Not that we should forget the good times of the past but we are going to spend the rest of our lives in the future!

[Sven]: Well, of course the present is not all that bad. First of all there are the academic consortia. Some of these may eventually suffer. As these corporate mergers continue, the number of sponsors will drop. There is some R & D going on within the contractors like Veritas, WesternGeco, and CGG. But these companies cannot possibly be expected to pick up and continue developing what the oil companies, in their short sightedness, have abandoned. They do not have the financial resources to do so. Take, for example, the former Amoco lab – shortly before it disappeared, it had operated with a budget of some $80 million a year. Now what contractor is going to have $80 million a year to spend on R&D, even if they wanted to?

[Sam]: But at whatever level, the technology will continue to develop. It may be brought from outside, it may be developed at a lower level inside the industry. But it will continue as problems evolve, and as we move more and more into the reservoirs. It will happen. It won’t happen the old familiar way, but it will happen.

[Sven]: Yes, I fully agree with Sam. We are simply not going to see a return to the days of large company-funded laboratory research programs. Perhaps if we become more aware of what people in related disciplines are doing, we might find the an alternate way forward.

[Larry]: There was an article in The Leading Edge a while back, addressing the alarming decrease in R&D effort within the oil and gas industry. It was either yours or Ken Larner’s.

[Sven]: We coauthored that article, along with Dave Lumley. Several other people in the industry also wrote articles addressing these issues. But unfortunately, who in the top management of oil companies reads The Leading Edge?

[Sam]: I think Ken went one step further and mailed one of his original articles to executives of many companies.

[Sven]: Actually, that was Marion Bone at a time when he was the SEG President. Ken and I, as well as others, convinced Marion Bone to send a letter to all the CEOs of all major oil companies at the time, pointing out the concern that R & D in the oil industry is was plummeting to a point of no return. The interesting thing was that not a single CEO of any oil company responded to that letter, which is kind of indicative that these gentlemen simply did not care, assuming that they even saw the letter. I imagine many of the letters never even reached the CEO’s desk. The letter came from the President of the SEG, and he never heard from anybody, not one.

[Satinder]: You have been teaching courses and going around giving talks. You must have been to many universities. What changes do you see in the academic level of the students between the 60s and the 70s and now?

[Sven]: Well, one change I see is that students are being better prepared. When I was a student, people knew very little about programming; there were no programming courses, nobody had been exposed to computers. Current generation students are more computer-savvy than the professors. So the roles have reversed. I had to learn computing from my professors. Nowadays, many professors often learn computing from their students. So that is a tremendous change.

[Satinder]: Apart from the training part that is going on in the universities, can you comment on the curriculum that is being followed?

[Sven]: It is difficult to generalize, because some universities are oriented more to the training in exploration geophysics, others are not. For example, the Colorado School of Mines in the States, and University of Calgary right here, not surprisingly tend to train primarily exploration-oriented geophysics. Other universities do that to a much lesser degree, for example very few exploration geophysicists come from Caltech and I could also say very few from my old school, MIT. Not necessarily a healthy development in my opinion. Some universities are more geared to exploration and some more to global geophysics.

Ideally, I think there should be a mixture of both. The separation is unnatural because after all, it is all geophysics. Global geophysics could benefit a lot from the techniques we have developed in exploration geophysics, but global geophysicists rarely come to our meetings. It is also equally true that we tend not to go to theirs. So there is a lot of room for improvement.

[Larry]: Actually, many of our students are getting jobs in the oil and gas industry, not many in R & D. But their technical background serves them well in those companies. Any advice you have for them?

[Sven]: Geophysicists pick up computing skills, and also study certain amount of geology and petroleum engineering. The big problem right now is not about communication between geologists and geophysicists. I think it has improved enormously in the past 40 – 50 years I have been in this field, but there is a much that needs to be done between the geophysicists and petroleum engineers. There is still a lot of misunderstanding about each other’s capabilities. There is all this talk of asset teams. The word sounds great but in reality there’s still a long way to go. I think one way to start the process is to train the students so that they can communicate not only with geologists but also with petroleum engineers.

[Sam]: Still, we are so much further down that path than we were.

[Larry]: Our students are very fortunate to be in a town where geophysicists can be employed and many of them especially our grad students. Almost 100 per cent are placed in these times. Of course there are years when they do not find jobs.

Photo 03
4 Doodletrainers – Brian Russel, Sam Gray, Larry Lines, Sven Treitel. Instructors for two day course: “Applications of Geophysical Inversion and Imaging”.

[Sven]: It is easy to advise students to become generalists, take the standard math, take a lot of computer science and so on, but the fact is that this is still a very specialized field. I mean, you cannot just get a grade degree in computer science and become an effective geophysicist. Geophysics is a field in its own right. It is not fair to the students to claim otherwise. There is always much to learn on the job. So you have to make a decision that you are interested in exploration geophysics and want to make a career out of it.

[Sven]: This brings me to another subject. That is, enter your career with open eyes and realize that you are not going to stay with the same company for long. I only had three employers in my entire career, but that I think is changing. My own sons, who are in computing, change every year or two.

[Sam]: And then people change industry also. People move into and out of the industry.

[Satinder]: Is it a good development?

[Sven]: I do not know if it is a good development or a bad one. It’s just different. One mistake you can make is to want to go back to the past.

[Satinder]: Sven, tell us about your most interesting professional experience.

[Sven]: When I think of it, one of my most interesting professional experiences was the one when our first attempts to apply predictive deconvolution actually worked. What made it so interesting was that I had been assured by some of our supervisors that it would not work. So there was a real challenge to get it to work.

[Satinder]: Were you at any time the President of the SEG?

[Sven]: No, but I did serve as editor for Geophysics.

[Satinder]: Then I can ask you this: A paper sent for publication in ‘Geophysics’ today takes something like one year to be published. Do you think there is any chance of improving the turn around time?

[Sven]: A lot could be done. First of all, they have just gone to the a new system, where manuscripts are submitted electronically and viewed electronically. While the system now is experiencing growing pains, I think the prospects for significant improvement in turn around time are growing. What it will be is anybody’s guess. The weakest link in the chain is still the willingness of the reviewer to return the review in a few weeks. While we have some reviewers who are that diligent, most are not. So you have to rely on the reviewer. An alternative that has been suggested, and I do not think it is a good alternative, is not to have anything reviewed at all, and publish everything that arrives online. I don’t think this is a good idea.

[Satinder]: Reviewing does improve the quality of the papers.

[Sven]: Right. I do not think I would like to read a journal that has publishes unreviewed papers. This way somebody has separated what needs to be read and what does not. This is admittedly a very imperfect way to separate what should be read from what might better remain unread.

[Satinder]: Separate the wheat from the chaff.

[Sven]: I was subjective, very admittedly, but you need some filter.

[Satinder]: I think the SEG has started doing something to encourage reviewing turn around times by giving cash vouchers to the reviewers.

[Sven]: Oh, yes, that approach is very nice. I like these vouchers myself. I spend them every year at the SEG. But I do not know how much it improves the turn around time. Then you have to take into account the fact that those of us remaining in the industry are under greater and greater pressure to deliver work because we are fewer and fewer in number. Many people asked to review just simply do not have the time to do the reviewing.

[Larry]: It is an interesting question. There are loose filter reviews in The Leading Edge and the RECORDER, and probably more people read those articles than the ones published in Geophysics. But for ‘academics’ it is important to have the peer review process; academics are given more credit for a paper published in Geophysics. Then there is a decision – do I submit a paper that will have a larger readership and be reviewed quicker or one that will give more points.

[Sven]: Just as an example, as I was coming up here, I was reading the latest issue of the The Leading Edge. I found there are some very interesting articles, but there are were also some/a few mistakes that should not have appeared. Those probably would not have appeared/occurred had there been a more careful reviewing process in place. Nevertheless, I am not really criticizing the The Leading Edge. The great advantage of TLE is they turn around so fast. A price has to be paid, and the price is the errors that are likely to creep into these articles.

[Satinder]: Do you get a copy of the RECORDER every month? Are you a member of CSEG?

[Sven]:No, sometimes Larry sends me copies.

[Satinder]: Given that you were the editor for Geophysics, I was hoping to get your comments on the editorial standard of the RECORDER. The Canadian Journal of Geophysics has been dormant for some while, so to a large extent the RECORDER is filling the void. We have put considerable effort into improving the quality, so it’s always valuable to get feedback from someone with your editorial experience, to hear if we’re on track.

[Sam]: I would say there is a huge difference in the quality in the last two years.

[Larry]: Yes, it’s vastly improved. (Pointing to a recent issue on Sam’s table) There is a list of contents for the latest one; in many respects it’s like the TLE. It is really a testimony to Satinder and Oliver that they have been able to bring it this far. There are some mathematical papers too in the RECORDER now. So we can call RECORDER our journal in some sense. It is a question of how many technical articles versus those that are descriptive that the readership wants. In your case Satinder, you have struck a nice balance here.

[Sam]: Yes, you have done an excellent job under the circumstances. You have a certain amount of pressure every month to get a number of articles and sometimes it is not so easy for you. I know that.

[Satinder]: We keep pestering people <laughter>!

[Sven]: You pester them until they write <laughter>, or you write your own! But And you have a large population of geoscientists who are right here in Calgary!

[Satinder]: We have really good people here. It is more a question of availability of time for them and when they find time they do write contributions.

Well, I would like to thank you Sven, Larry and Sam for sparing time for this excellent conversation we had.

Sven: Your job now is to edit out the chaff <laughter>.

Satinder: Yes, I will make sure you see lots of wheat <laughter>.

Sven: Thanks for asking me. I do not often get interviewed.


Share This Interview