An interview with John Castagna

[Satinder]: John tell us something about your background, what all you did, where you have worked and what you are doing now?

I did my Ph.D. at the University of Texas with Milo Backus and from there went straight to Atlantic Richfield Corporation (ARCO). I was with them for about 16 years before I moved to The University of Oklahoma. I have been there for the past 6 years, continued my research there and put in a lot of effort into building a consortium and a research institute. In the past few years, in order to make it easier to generate funds I was influenced to start a small contracting company. Its easier to bring certain kinds of funds into a consultancy rather than to a University.

[Satinder]: What inspired you to take up geophysics as a career?

Well it is bit of a long story. First I need to explain why I took up geology as a career. As an undergraduate I went to a commuter college in New York City, where 90% of the students who went to sciences were Pre-Meds. My initial intention was to major in Chemistry. After I received a C in the second term of Organic Chemistry despite a 95% average, I decided that maybe I pursue something different. At that time I was also taking a Geology class that I enjoyed very much. As a kid I had always had a strong interest in collecting rocks, minerals, and fossils. The fact that there was an oil embargo at that time suggested to me that it could potentially be a lucrative field. So, I became inspired and pursued that. Geophysics came later, primarily through the influence of Peter Bell, from the Carnegie Institute of Geophysics in Washington. He got me interested in solid-earth geophysics, and from there it was a small step to exploration geophysics.

[Satinder]: Well now that you are on the other side, how do you inspire your students to take up Geophysics?

Well, to tell you the truth, I do not try to attract students into Geophysics. Having worked in the field for as long as I have, I can say that it has been a wonderfully challenging career and continues to be. However, given the uncertainties in the industry, especially in the States and the way personnel are treated by oil companies, I feel reluctant to try and attract people to this field. So if someone comes to me and says I am committed to Geophysics, my life is Geophysics, and there is nothing you can do to dissuade me from pursuing Geophysics, then I take them on. Then I do everything I can to make them good geophysicists. I have got 20 of them right now. Everyone I know who is truly committed to this profession has had a wonderful career in geophysics. People who had thought they were going to work 9 a.m. to 5 p.m. and have a fat secure salary without giving it their all, they have not really survived in geophysics.

[Satinder]: What prompted you to quit your job at ARCO and take up teaching/research as a profession?

I enjoy teaching as I get a lot of satisfaction out of that but my true love is research. When I went to ARCO, I joined ARCO Research. I then moved into operations, and that was a lot of fun working as an explorationist and in field development, but I could never cure myself of the desire to do research. However, from the time I started working for a major oil company in the early 80s to the mid 90s, the research climate really changed. I realized that a university would provide me the security I wanted and that is what I obtained at the University of Oklahoma with a tenured chair position. So I can research what I want to, I can say what I want to, politically correct or incorrect. I cannot lose my job for having the wrong opinions – for example thinking that runsumming seismic traces is not always the optimal thing to do. So I have the freedom to pursue geophysics the way I want to without worrying about the current brainless party line.

[Jon]: Do you have the same resources that you had within ARCO in the university environment, to do the type of research you want to?

No, the resources are not the same. The biggest problem is the paucity of data. However, we do have a research consortium – thanks to the generosity of many companies, we do have some variety of datasets to work on – but I am never satisfied. Through the commercial companies and the consulting that I do, I get exposed to the real problems and limitations people are dealing with. That is important. The computing resources that I have access to are not equivalent to what I would have within a major oil company, but computing has advanced so much over the years that compared to what I was used to 10 years back, I have plenty of computing power. Limitations are sometimes good because they force you to be clever rather than use brute force.

[Satinder]: What personal attributes helped you achieve the professional status that you enjoy today?

I would say one of my biggest strengths was that I was trained as a geologist. That makes me unusual for a research geophysicist. That gives me a much different perspective of things. It is funny that one of the things I enjoy doing most is developing algorithms. Coming from a geological background though, my approach tends to be different and I tend to be fairly independent in my thinking.

[Satinder]: During your building years who were your mentors and why did you admire them?

The first and foremost is Milo Backus at the University of Texas. Once I decided to pursue Geophysics after having done a masters in Geology, my major thinking was that if I were to get into the oil industry, I should be going to a school sitting right on top of the stuff. I did not even know who Milo Backus was. That is often the case. When graduate students select a school, they do not necessarily know the individual professors. That happens a surprising number of times. I did realize later, how incredibly lucky I was to be associated with a man like that. I think the gifts he gave me I cannot even begin to enumerate. He taught me to look at the data first and to salivate when the theoreticians say something can’t be done. Certainly he would be number one. Peter Bell was definitely important for my decision to get into Geophysics. And then my journeys into rock physics were largely due to the influence of Mike Batzle. He was a mentor for me in my early years at ARCO and he never kicked me out of his office when I incessantly bounced my ideas off him. The work we did together was always bootlegged and extracurricular but I think time has shown that it was worthwhile.

[Satinder]: Now what are your aspirations for the future?

My aspirations for the future are to continue doing research and be able to stabilize research funding for the group of people I work with. My ideas tend to be such, that they are too many or grandiose, so I cannot implement them myself and really need to be working with a group of people. I really enjoy the situation where you have a number of specialists who are expert in areas that I can never hope to be expert at. We could work together and create something that is bigger than any of us could generate individually. But in order to be able to do that you have got to finance that kind of operation, you’ve got to pay salaries, overheads, computing equipment, etc. My aspiration for the future would be to solidify that kind of funding either through endowments or other means.

[Satinder]: How do you manage your time between your teaching, research, your consulting, your position with SEG sometime back, etc?

What I found out was, once you get past using all of your waking moments you can continue to take on more work and it does not affect your work load, because you do not have any more time. It is not a problem at all. You just use up all your time and solve problems while you are dreaming.

[Satinder]: Does it allow you to strike a balance between the different things you wish to accomplish?

just do what I want to at the moment and in retrospect, that system has worked pretty well for me.

[Satinder]: Interesting. Apart from AVO what other areas in geophysics do you like and what are your contributions there?

I guess I have to say I love all aspects of seismic reflection; I have been able to filter out any knowledge or expertise that I might possibly have in non-seismic methods. I am amazed at the filter I have. I am able to forget the simplest of equations when it comes to magnetotellurics for example. I do focus on seismic. I have always been oriented towards the seismic interpretation side because of my geological background. But as opposed to geological seismic interpretation, I like to focus on the area you would call seismic analysis and it really has to do with how can we extract more information from seismic data? Other than the structural or stratigraphic information how can we make inferences about lithology, porosity, pore fluid content and how can we bring together various kinds of a priori information available to allow one to do an interpretation like that? So that was always part of a larger interpretation, which inevitably involves the utilization of seismic inversion. I think seismic inversion has to some extent gotten a bad name in the industry. In my opinion, it is often badly used. Inversion has to be viewed as an interpretation tool rather than as a processing algorithm. When viewed that way you can do fabulous things with it.

[Jon]: In the past we have talked about AVO as an anomaly identification technique, more so, than trying to invert for elastic parameters. Where do you think we sit with AVO? Do you think we can actually invert for elastic parameters and get meaningful estimates out of that?

You guys are already doing that I think. I have always been cognizant of the non-uniqueness of the situation. But as long as you go in with that knowledge and worry about things like amplitude calibration you can do wonderful things. If you expect to just be able to take a seismic gather, put it into a black box and get the answer out, you are going to get burnt. But if you recognize that what you are trying to do is find interpretations that are compatible with your data or constrain your interpretation and use seismic inversion as a tool to do that, it can be used very effectively.

[Satinder]: Ever since Ostrander introduced the interpretation of AVO gather, how do you think the perception or expectation from AVO has changed over the last 18 years or so?

Oh jees! That is a story. <laughter> Actually I had started doing a study of using AVO as an extension of the whole larger issue of introduction of new techniques into the industry. I started writing something up very scholarly, but never got time to pursue it. But AVO has got an interesting history. Many people were doing gather amplitude interpretation in the late 70s. Maybe it was not talked about, but many people were doing it. I was introduced to AVO in 1978 as a graduate student at UT doing a project with Clark Wilson and then later with Milo Backus. Especially after AVO was introduced by Ostrander at Chevron, it was initially followed by a euphoric period. It was very similar to the euphoric period after bright spots were introduced. People went out and started hunting for pure anomalies without understanding what they were looking for or what the problems and pitfalls were. We were doing some moronic things – operating on rules of thumb or myths that just weren’t valid. I have been complaining for a long time that interpreters were just looking for amplitude increases with offset and then using attributes based on amplitude increase with offset. They needed to be a lot more careful about what they were looking at, what they were calling anomalies. As a result that caused all sorts of disillusionment and dissatisfaction. So by the late 80s AVO was almost an undesirable word and some companies had to change the name to gain internal acceptance. SEG actually had a research workshop in 1992 on ‘How Useful is AVO analysis?’ I mean it was still an issue whether or not it was useful in some quarters. All along there were some workers who were using AVO very effectively and happily, while entire organizations including some top oil company management felt AVO just did not work. My response was that the laws of physics work – your employees just do not know how to use the technique. It is not just a matter of AVO not working. A lot of good approaches were developed. More importantly, people recognized the non-uniqueness of the responses; the different kinds of responses to expect. Now it has gone to a point where we quantitatively do AVO. When I wrote my parts and edited the book, ‘Offset–Dependent Reflectivity – Theory and Practice of AVO Analysis’, which came out in 1993 - the papers were actually written before 1990 – when AVO was not commonly being done quantitatively. It is now much more common.

[Bill]: I was one of those you mention who spent most of the ’80s ignoring AVO -but I did buy your book- not because I did not believe in it but because I thought at that time it was too difficult to use and just did not want to go there. However, my impression through the same period was that interpreters were drilling bright spots, dim spots, and spots of any intermediate intensity and so we had to deal with amplitude interpretation whether post or pre-stack. There came a point when we had been drilling the bright spots, without success, so we started to drill more of the dim spots. Because of this confusion in post stack amplitude interpretation I decided I needed to understand and exploit pre-stack AVO. Even at the general session of the EAGE this year, Andy Wood, head of the global exploration, Shell international E & P, mentioned publicly that DHI’s, I presume it was AVO, did not significantly alter the actual success rate of prospects drilled in recent years. I am interested in your opinion on this.

I have worked with a lot of ex-Shell people and Shell has historically been a huge believer in DHI’s. I know the years when Shell attributed the bulk of its success in the Gulf of Mexico, for example, to DHI’s. As a matter of fact, I was privileged to work with M. Downy for 3 years at the University of Oklahoma (OU). We both came to OU at the same time. One of the things Marlan did when he became president of ARCO International was to insist that the interpreters be cognizant of the different kinds of DHI’s. You could not just ignore them. They had to be taken into consideration. Tom Velecca, another ex-Shell guy who was exploration vice-president at ARCO made the point that AVO was there, it was part of the seismic signal you were seeing and you are not doing your job if you are not understanding the whole seismic signal. Recently I have had the pleasure of working with Mike Forrest. So my understanding of the Shell approach is that understanding of rock properties and post-stack and pre-stack amplitudes is a fundamental part of doing the job and it surprises me that someone from Shell comes along and says there is no influence on the success ratio. I suspect that when the global statistics are dissected one will find that there are places in the world that you simply don’t drill a well without a DHI and that these cases are being diluted by many inappropriate utilizations under circumstances that are not conducive to DHI, or more specifically AVO, analysis. I hope that these comments are not part of a general dumbing-down trend. That would be a shame.

[Jon]: One of the issues in terms of AVO reliability is how you prepare the data. Do you have any thoughts on what would be a good processing sequence for AVO or there are some big issues you should look out for while considering doing an AVO study?

The larger issue is to try to preserve or recover the amplitude variations across the gather, which are truly affected by reflection coefficients. In order to do that requires that your processing does not corrupt the information that is there. One of the common problems that I have encountered is non-amplitude preserving processing procedures. And this could go back a long way in the processing flow. There are so many different processing packages around that it is difficult to keep track of every algorithm out there and, unless you have done the processing internally with algorithms you know, you have to assume that you could get hit by any process at any point. For example, when we first started doing AVO at ARCO we found that our trace by trace decon was not amplitude preserving because nobody had bothered to make sure that the operators had constant energy across the gather, because they were not worried about AVO when they designed this way back; particularly, the imaging algorithms. So, I basically take it as given when AVO data is provided to me by a routine processing shop that amplitudes have not been calibrated. I try to make sure they have done nothing to harm the amplitudes and then I believe the amplitude calibration is part of the interpretation.

[Jon]: What do you mean by amplitude calibration?

In other words if you take a raw gather that has not had amplitudes corrected the amplitudes are going to vary in ways that are not directly related to the way reflection coefficients are varying. There is going to be something undesirable variation superimposed on top of the reflection coefficient variation with offset that needs to be corrected for. So how is this correction done? You can try deterministically to figure out everything that has happened to the amplitudes as the signal has propagated down to the target and back and try to back these effects out. To my mind, that is extremely tough; you may be able to do a ballpark correction. But I have not had the courage to state that we can correct for everything that has happened all the way down to the target. One thing that is required to do this that is generally not well known is the geological model down to the target. If I knew that, I could do a lot of things. So ultimately, you need to depend on some kind of a statistical approach or some compensation type of approach and that is very much dependent on the problem you are dealing with and the geological situation. If you are very lucky, you could find a reference reflection to compare to but very often you cannot do that. Sometimes you try some kind of windowed equalization, even resorting to a long window AGC, an inappropriate thing to do, but sometimes the best of poor options. But ultimately your amplitudes versus offset, especially where you do not have pay, by the time you are done with this procedure, should be acting the way your synthetic model of the primary coefficients only varies with offset if you have done your amplitude calibration properly. So that is an issue. Obviously, signal-to-noise ratio is an issue. You do not have your greatest weapon, stacking. Since you are working with pre-stack data, you have to try and hit it with pre-stack noise attenuation techniques, like radon or other approaches. For that you need sufficiently fine sampling and sufficiently effective fold, i.e. fold within your mute.

[Bill]: I was at the special session at the SEG where you introduced the type 4 anomalies. In the last year I noticed a converse to type 4 where there is a false type 1,i.e. if we have a high impedance interval that has a decrease in amplitude with offset at its top, this can in fact occur from an increase, not the expected decrease, in Vp/Vs or Poisson’s ratio. I was wondering if you had warned or told people about this possibility.

When you say abnormally high Poisson’s ratio you are saying this is not a reservoir but looks like a type 1.

[Bill]: Yes, the opposed quadrant from a type 4 on an Rp-Rs cross plot, because as the gradient is Rp-2Rs, it does not match the Rp=Rs line. So on the opposite Rp polarity side or quadrant from your type 4 you have a high impedance, positive Rp top with amplitudes that still decrease with offset. So it looks like a standard type 1. But it is a high impedance interval with a higher not lower, Vp/Vs ratio to the background. So it is false type 1.

No, I have not warned people about it specifically

[Bill]: Because of your introduction of type 4 and this possible pitfall with false type 1’s, I have got away from offset based amplitude rules now, which is where we get into inversions. There is yet another confusing type, the dead flat amplitude with offset type i.e. in transition from type 3 to 4, which by definition has no variation and we miss targets like that. Because guys are looking for some variation and this by definition is strictly not AVO.

Well, this false type 1, I have not warned people about it specifically, but what I have warned about is having the correct AVO polarity. It has to do with deviation from a background of some kind. You have a certain deviation usually the way we plot things below the trend or depending on the way we plot, it could be either side of the trend. An abnormally high Poisson’s ratio will move you to the other side of the trend. So, if you have a layer which is a gas reservoir your anomaly should be an abnormally low Poisson’s ratio response over a high Poisson ratio response. So a low/high is analogous to when you are looking for low impedance reservoir on a stack when you are looking for trough over peak. When you are looking for bright spots, we want to see that leading trough. So we do warn people about that. There are lots of false AVO anomalies that are the other way round.

[Bill]: With your stature in the business, I am surprised that you have problems getting data.

Often the problem is getting the data published.

[Bill]: So you can work on interesting problems, which may not necessarily be published.

We do a lot of proprietary work. That is one of the ways we make ends meet. We work to subsidize our research, for the right to do research. I call it paying the rent. It is something I did early on at ARCO. I gave management what they wanted and did the research I wanted to on evenings and weekends. Later in my career I got more stubborn. I see a lot of proprietary data but getting data released for publication is sometimes difficult. Sometimes there are partner issues, spec data, etc.

[Jon]: There has been a proliferation of attributes and different ways of interpreting data. Do you have any favoured methodology?

I like all the attributes as long as the attribute set you use has not lost information. An example is using only the AVO intercept times gradient product indicator. An attribute like that will cause you to lose information. As long as an attribute does not erase information that is there, as long as you interpret it in an appropriate way, and it shows anomalous behavior, it is good. The additional advantage of different attributes is looking at things with different colour glasses on. It might be that the same information is there in two different attributes, but it is just the way one happens to plot on a given color scale or the way one happens to stretch or contract on an axis that could make things apparent, that might not have been noticed otherwise. Seeing things from different perspectives, taking a number of different views at the same information sometimes is very helpful. So I do not have a favoured attribute, I like them all.

[Bill]: There is a lot of discussion about getting reliable density information from AVO. What are your opinions on this potentially controversial claim?

I will first tell you what the theory says and then I could tell you what I observed. For many of the density term extractions that you see talked about or shown, you are not necessarily told how that attribute is arrived at. I have to guess what is being done. The approach we tried early on in the 1980s was to go to the 3rd term in the Shuey’s approximation, which is related to the contrast in the P-wave velocity. You subtract that from your normal incidence P-wave reflection coefficient. You are left with the contrast in density. What we found early on was the signal-to-noise ratio is a hundred times worse than your stack. So it’s very noisy attribute and you have to interpret that attribute with that in mind. When I see a density map presented, one thing that I would love to see, but do not get to see in press, is the standard error of the estimate of that attribute. I have asked this of a number of attributes that are presented. Look, you are providing a C term, just provide a standard error of the estimate. I have never seen anyone display that. One of the reasons is it is so wide; it could allow anything in there. From a theoretical standpoint, I am really suspicious. According to my training, this is time to salivate.

From a practical standpoint I have seen some of the most incredible examples published that are very very convincing. So what it suggests to me is that there is something else going on, which is influencing it or biasing the extraction of this attribute that is causing it to be meaningful in physical terms when the theory is telling us it should be tough. Even from synthetic data, adding random noise to synthetic data, the error in the 3rd term is high. So if you throw in anisotropic effects getting more difficult at far offsets etc., it causes me a lot of theoretical concern, but at the same time I cannot deny what I am seeing.

[Bill]: Jon has written some recent papers on the sensitivity and the error bars associated with a number of attributes.

[Jon]: I agree with everything you said, I think the 3rd term has a large uncertainty associated with it, but maybe in some situations where there are wide enough angles and a large enough signal-to-noise ratio, we might be able to get something useable. I also agree you need to have quality controls to be able to see that. In doing this for large angles you have to worry about supercritical angles. Do you have any thoughts about how you can identify supercritical reflections and prevent them from going into the analysis?

When I was extracting this attribute I was trying to do an angle mute. I was trying to cut out larger angles. In terms of identifying and seeing larger angles directly from the data, sometimes you can see the critical refractions. Maybe I fooled myself, especially on a strong reflection. The other thing that you sometimes observe is that going to the theoretical plane wave model, you expect a very sharp peak at the critical angle. But with real data it is not so clear. What you see is a gentle roll over. Now there are many things that could cause a gentle roll over in your AVO. If it is specific to an event, events above and below are not having this unexpected loss of amplitude, that in my mind could be caused by a critical reflection.

[Bill]: What about the phase change associated with supercritical reflections?

The phase change, depending on how large it is, could be mistaken for anisotropy. It often looks like a moveout error. If the processors have done too good a job of flattening the gathers, maybe you won’t notice it. But you are right. If things are properly processed, the moveouts are done above and below, you will see what looks like an anisotropic effect on only the target horizon and you will be able to recognize that.

[Jon]: One of the things I am curious about, is when you move into the Gulf coast, a strong velocity gradient would give apparent anisotropy. The intrinsic anisotropy will be different. What do you do to model such a situation or do an AVO extraction? Do you have to worry about apparent versus intrinsic anisotropy?

Are we talking about kinematics or amplitudes?

[Jon]: That’s my question.

For kinematics it does not really matter for AVO purposes. For the amplitudes it is common to ignore anisotropic effects. Not everybody is paying as much attention to amplitudes of anisotropy or anisotropy effects on reflection coefficients as they should be.

[Jon]: So people are assuming some sort of 3-term model, like the Shuey model and doing the extraction.

[Bill]: There are a number of papers trying to link the kinematics to AVO and I find that very interesting. Do you think it could be a more routine process? In other words if you were picking kinematic information -with or without anisotropy- from velocity analyses on a reflector and had the same reflector’s instantaneous AVO inversion information, then there would have to be some convergence with the AVO backing up the move-out.

It is a tough problem, because kinematics is more of bulk kind of measurement. The amplitudes are more of a localized thing. It would be similar to using interval velocity measured from velocity analysis and then using that to constrain a reflectivity inversion. I know of one research group that is looking at anisotropic AVO inversion. I think that the way to handle it is both at the same time.

[Bill]: So you think there would be tools in the contractors’ hands where they would be tempted to pick a certain NMO correction and the AVO would say, no no it is this pick, instead. Without full-blown AVO estimation of interval parameters there are some tools that give reasonable interval velocities after pre-stack time migration with continuous interval velocity updating. From this we see big discrepancies between inverted AVO information compared to interval velocities. It would be good if the person who has to do both has a tool that some how converges on a more consistent solution.

I agree.

[Bill]: Do you see a lot of AVO being used for pore pressure prediction?

Yes, indirectly, it was a routine standard procedure at ARCO to use AVO–derived velocities to do pore pressure prediction. As a matter of fact, some of my buddies, Swan, and Kan, have published papers in Geophysics recently.

[Bill]: I think this is ignored in general in that part of our industry. In my company, when talking to geologists involved in pore pressure prediction and even from a logging point of view, I suggest they look at shear logs as well for better lithology control. They are usually basing everything on the P-sonic logs for coming up with gradients or trends.

[Satinder]: You mentioned about anisotropy. There was Bill’s paper in the Special Issue of the RECORDER where he shows examples of fizz gas showing up as an anomaly. How much AVO are we using effectively in the industry today for both these effects?

I think there is a relationship between anisotropy and gas saturation in the reservoir, especially in the material above and below the reservoir. That is something that can produce effects that people would like to inspect in general.

[Bill]: In the Gassmann’s equation, though I missed the Calgary SEG workshop ‘Beyond Gassmann," the assumption is that rigidity is less affected or unaffected. Do you think we have to get beyond that?

At the sonic frequencies we have to go beyond that. The traditional view was that sonic logs are in the low frequency range and I am of the opinion that is not always the case. Sometimes the sonic log is in the transition to high frequency range and where if the pore pressure does not have time to equilibrate between pores as the wave passes, then the shear modulus is going to be dependent on gas saturation, and that certainly happens with theoretical modeling. It is easy to understand why that might happen in a rock that does not have good permeability. I personally feel that some effects that have been attributed to patchy saturation are eventually going to be recognized as being due to dispersion of this kind related to pore interconnectivity and the distribution of gas between pores in low quality reservoir rocks.

[Bill]: When is your next book on AVO coming?

If this is going to be a scientific or empirical answer, it is never. Because in 1996 I proposed the next book and had an outline written. If we project the progress on the book over the past 5 years and carry that forward, the answer is never.

[Satinder]: Being on the optimistic side?

Probably when I retire.

[Satinder]: What message do you have for fresh graduates entering our industry?

Eat, live and breathe geophysics. I think everybody I know that is totally committed to the profession may have had disruptions in their careers but they landed on their feet really well. I think it is an incredibly fun and challenging career.

[Satinder]: John, my buddy Oliver, who is the technical editor for the RECORDER and could not make it here today, asked me to specifically ask you, what do you do to get funding for research at your institute. What are the different ways and means that you adopt?

We apply for many grants to various funding agencies. We spend a lot of time writing grant proposals, which becomes very tiring because a small percentage of them get funded and I would rather be doing research than writing proposals. Then we go and beg at oil companies. Fortunately, there are enough people out there that feel that the contributions we have made have been worthwhile and support our research. But I have to say that to a large extent these companies are being good corporate citizens. They know the money is being spent to generally elevate technology, to improve communication and technology utilization, and to educate students, but they are not seeing direct benefits that can be easily quantified this quarter. Those kinds of contributions are really appreciated. Then there is a lot of proprietary work. I am finding it very easy to get funding of this kind. We are getting to the point where the majority of our funding is in that form and so we use that to enable our research.

[Satinder]: Well, John it has been nice talking to you, knowing more about you and the work you are engaged in. I thank you very much for giving us this opportunity.

Bill and Jon, I thank you for taking the time to come here and join in this discussion.