Geothermal seismology considers the seismological ramifications of injecting fluid into crustal rocks for the purposes of geothermal capture or carbon sequestration. Bruce Julian, a world-leading expert in this field, developed this specialty by way of a long career at the United States Geological Survey, where he pursued a broad range of research projects on geophysical inverse theory and seismic wave propagation, and tectonics. Julian also pioneered computer programs and their applications toward devising ever more accurate capture and analysis of seismological data.
Having spent the entirety of the 1960s at Caltech, first as an undergraduate in engineering and then at the Seismo Lab, Julian is well positioned to provide perspective on the Lab's classical early era, and its transition amid the plate tectonic revolution toward a more modern integration with geophysics writ large. His thesis focused on the three-dimensional structure of the mantle, and for his postdoc at MIT, Julian worked on seismic discrimination; i.e., determining the source of seismic activity, which can result from earthquakes or other natural processes, and can also occur from large scale explosions such as the detonation of nuclear weapons.
In retirement, Julian works as a consultant in the geothermal seismology field. As the need for both geothermal energy and carbon sequestration grows - both owing to climate change mitigation, so to will demand for expertise in this area.
DAVID ZIERLER: This is David Zierler, Director of the Caltech Heritage Project. It's Wednesday, May 16, 2022. I'm delighted to be here with Dr. Bruce R. Julian. Bruce, it's great to be with you. Thanks so much for joining me.
BRUCE JULIAN: Thank you very much.
ZIERLER: To start, would you please tell me your title and institutional affiliation?
JULIAN: I was a Geophysicist at the US Geological Survey for 35 years, but I officially retired about 10 years ago. I'm still doing some geophysics and seismology, working with my partner, Gillian Foulger, who has a little company called Foulger Consulting, and we monitor earthquakes in geothermal areas and do other things of that sort on a very small scale.
ZIERLER: What are the kinds of clients you have in your consulting work?
JULIAN: Mostly geothermal companies. There are several geothermal power plants in the western US. If you drill holes in the ground and take steam out or pump water in to fracture the rocks, you tend to cause earthquakes. This is something I got into later in my career. But the clients are often geothermal companies generating electricity by taking hot water or steam out of the ground and running turbines to generate electricity. That tends to disturb the stresses in the rocks and causes earthquakes. And they're required by government agencies to keep track of them in the hopes of avoiding doing something dangerous.
ZIERLER: What have been some of the major areas of research you've worked on over the course of your career at the USGS?
JULIAN: At Caltech, Don Anderson started me working on the question of three-dimensional variations in the upper mantle, and much of my career has been working on that. How to trace rays, how to see the three-dimensional structure of the interior of the Earth, and how to infer that structure from seismological observations.
ZIERLER: Tell me about some of the major institutional collaborations you've had with the USGS. Let's start with Earthquake Processes and Probabilities Megaproject. How did that get started?
JULIAN: That's a lot of bureaucratic mumbo jumbo. At Caltech, Don started me working on the three-dimensional structure of the mantle, and that's what my thesis ended up being about. I then went to MIT as a post-doc and very soon was hired by Lincoln Lab at MIT, which was studying what they called "seismic discrimination", jargon for how to tell the difference between earthquakes and explosions. Because a very important application of seismology was (and is) to try to learn how to monitor a potential nuclear test ban treaty and it was very confused by the fact that the Earth has lots and lots of earthquakes, and they look quite a lot like explosions on seismograms. It was a big project for many, many years to upgrade seismology to better be able to tell the difference between earthquakes and explosions. Major tools used for this purpose at the time were large seismometer arrays. These were patterned after radar arrays, and can be used to detect weak signals by processing the outputs from large numbers (hundreds) of distinct sensors deployed in a pattern over an area. But they also provide direct evidence about the heterogeneity of the propagation medium, the earth in this case, and made it clear that this heterogeneity had to be accounted for in the seismic discrimination problem. So this dovetailed nicely with the subject of my thesis, and extended it from North American upper mantle to the entire mantle. I worked on seismic discrimination for five years at MIT, then joined the US Geological Survey and moved to Colorado and then five years later to California. In California, the USGS was doing research on earthquakes in the hopes of predicting them. That was the main thing they were trying to do. I would say it never worked. I was still studying how to determine the structure of the earth and how to use that information to locate and identify earthquakes, and then got diverted into studying volcanic earthquakes.
About the time I got to California, there were some large earthquakes in California at Mammoth Lakes, which is situated in Long Valley Caldera, a volcanic area on the eastern side of the Sierra Nevada. These earthquakes were very strange in all sorts of respects. No one expected them. Instead of being a typical situation where there's a large earthquake and then aftershocks, there was a large earthquake, and then a day later, another large earthquake, then another large earthquake, then another large earthquake. There were four large earthquakes. It wasn't a main shock-aftershock sequence at all. And when people tried to study the source mechanisms of these earthquakes, they couldn't get a sensible answer. I remember one paper given at a scientific meeting where somebody showed all the seismic data relevant to the largest earthquake. There were hundreds of data, from networks in California and Nevada. This person said, "I just threw out all the data from Nevada because I couldn't get a double-couple answer." That means an answer that corresponds to pure shear slip on a fault, just two rocks sliding past one another.
That's always been the official theory for what causes earthquakes. And this person was doing this crazy thing. They said that half the data have to be thrown away because they didn't support the right kind of answer. It should've been a wake-up moment. When you don't understand something, that's great. There's something you can do there. I suspected that the motion on a fault didn't have to be only simple shear, that cracks could perhaps open or close, or that faults could be curved. There were various other possibilities. The rocks could be anisotropic (their properties could be different in different directions). In fact, all rocks are anisotropic, at least to some extent. I wrote an article in Nature showing that the Mammoth earthquake data were perfectly consistent with a more general family of source processes, and soon got a letter from Gillian Foulger, an English and Icelandic graduate student, who had found lots of similar earthquakes with apparently anomalous mechanisms in Iceland. So we got into studying what came to be called non-double-couple earthquakes, and particularly geothermal earthquakes. Volcanoes and geothermal areas have high fluid pressures. That's why they are economically important: we can use the high pressure fluids to drive turbines and generate electricity. High fluid pressure means there are likely to be cracks being forced open or closing, something more than just the simple sliding of two rocks past one another.
ZIERLER: Is volcano research a big part of what the USGS does?
JULIAN: Yes, indeed. The USGS has had a volcano observatory at Kilauea, on Hawaii, for several decades, and has now established similar observatories for Long Valley caldera, the Cascade range, and Alaska, and collaborates with volcanologists from many other countries. My office ended up being moved for several years into the building where all the vulcanologists were. A lot of the work was geologically oriented, with people mapping volcanoes and figuring out what flows and eruptions have occurred, when they were and how big and so on, but volcanology has become highly cross-disciplinary.
ZIERLER: Just to set the stage, before you got to Caltech, even in high school, were you interested in earthquakes and geophysics?
JULIAN: Yes, I was, but I had many interests. My family did a lot of hiking and camping in the Sierra Nevada, so I was interested in the geology you could see around you and questions like, "Why is the Sierra Nevada there?" But I was basically an electrical engineer. My father was an electrical engineer, and we had a garage full of electrical junk, radios and whatnot. And that was the main subject I was interested in. That's my undergraduate major at Caltech in. By the end of my undergraduate career, I decided to switch to geophysics because I'd taken a course from Frank Press in my junior year, and that led to a summer job, and I decided to switch to Earth science for graduate school. The Earth sciences were at that time, and I think probably still are, encouraging people from other disciplines to apply and I realized that my electrical engineering education was very relevant to studying other kinds of wave propagation.
ZIERLER: Was it a unique thing for Frank Press to be able to do something like teach an undergraduate course?
JULIAN: No, I think he taught it regularly. It was called Introduction to Geophysics. Intended, I think, to be a "cultural" course, for people from other majors to take. And several other Seismo Lab people taught more specialized undergraduate courses, aimed at Earth-science majors.
ZIERLER: What was the job this led to?
JULIAN: I was assistant to Don Anderson, who was a brand-new PhD himself at Caltech. He was interested in the three-dimensional structure of the mantle. People were using travel times of seismic waves to study structure of the crust and upper mantle. You'd have an earthquake, and if you knew where it was, then you could use seismograms recorded somewhere else to measure the time it took for the seismic wave to get from here to there. And the further away you got, the deeper the waves would go, so they would be traveling through rocks that typically had higher wave speeds. And you could unscramble the observed travel times to figure out how wave speed varies with depth. But the only theory available at the time was for one-dimensional models, with the Earth's properties varying only with depth. First off, Don thought that they probably varied in three dimensions, which was at that time a surprisingly radical idea, in the eyes of many geophysicists. But in fact, it's not a simple curve. It's not just a single line. It goes back and forth on itself. It has what are called triplications, which means that a lot of the relevant waves aren't the first-arriving waves. When you try to measure the travel times of waves, you tended to pick the first thing you could see on a seismogram and ignore all those later squiggles. But those wiggles are equally important. That was one of the main things Don put me onto, figuring out how to compute travel times for when the travel time curves have triplications, are complicated, and when the structure is three-dimensional.
ZIERLER: Tell me about working in the old Seismo mansion.
JULIAN: That was an amazing place. I believe Charles Richter and perhaps Gutenberg were responsible for getting that building and setting it up that way. It was a huge two-story mansion in the San Rafael Hills above the Rose Bowl. It was the last place you would imagine to be a laboratory. Everybody's office was really just somebody's bedroom or a kitchen or something once upon a time. The graduate students were in a huge ballroom, into which a dozen desks had been placed. Don Anderson's office was a former bedroom with an en suite bathroom, and he used the bathtub to store reprints. [Laugh] But one of the most amazing things about it was, every day, there were a couple of coffee breaks in a room in the basement where they stored seismograms. It didn't have a paved floor or anything, it was just dirt. It was an unfinished basement. Which was really an ideal, it turned out, kind of environment. Everybody was extremely informal, everybody was there, the secretaries and everybody else. You weren't talking to the people you usually talked to. You might be talking to some of the electrical engineers who built seismometers and things like that. And people might learn that somebody else actually knew the answer to the thing you were wondering. But you probably wouldn't have ended up finding that out without talking to that person. This very unconventional building really led to that happening a lot.
My father had worked at MIT on radar during World War II, in a place called Building 24. I found this out later when I was working there, and he came and visited. This building had been designed in a panic at the beginning of World War II. They had taken some Army plans for barracks and just said, "We'll stick these barracks next to each other and nail them together." It was a completely unconventional building. One of its properties apparently was that you'd get lost in it, so people would wander into other people's offices and say, "Oh, I thought this was my office." "No, it's my office. But who are you?" Apparently, it led to a very productive environment just because it wasn't what you expected an institution of that kind to look like, and it led to people running into others they wouldn't ordinarily meet. When the Seismo Lab was moved to the campus in the 1970s, Don tried to recreate an environment like that in the new Mudd Building. I never worked there, this was after my time, so I don't know how successful it was. But he tried to get the architects to build it so it was crazy and had bends in the corridor where you wouldn't expect them, and things like that. It had a large room intended for coffee breaks every day, which was intended to get everybody talking to each other. But it was fancy, with walnut paneling and no dirt floor. I don't think it worked as well, from what I've heard.
ZIERLER: Did Don end up being your thesis advisor?
JULIAN: Yes, he did.
ZIERLER: What other faculty at the Seismo Lab did you work with?
JULIAN: Charles Archambeau, notably. He was a very mathematically oriented seismologist, and was interested in the three-dimensional structure of the mantle. We spent many happy night-time hours in the computer center waiting for our latest output to come out of the machine. Stewart Smith was there, who had built one of the first digital seismographs in the world and also made the first observations of free oscillations of the Earth, from the great 1960 Chilean earthquake. Frank Press left and went to MIT about 1965, but I took his course on elastic-wave propagation during my first graduate year. It was one of the most valuable courses I ever took. I also talked several times with Jim Brune about earthquakes and how they worked. Keiiti Aki was there when I first worked as an undergraduate, but only for a few months. I got to know him better later at MIT. He was extremely inquisitive and original, and interested in many things, including volcanic earthquakes, earthquake mechanisms, and non-double-couple earthquakes. He died prematurely, of a head injury, which was a great tragedy.
ZIERLER: Tell me how you developed your thesis topic.
Don Anderson and Nuclear Explosions
JULIAN: One of the things Don had me doing at first, in addition to computing travel-time curves, was looking at seismograms from nuclear explosions. At the time, the US was testing lots of nuclear explosions underground in Nevada and occasionally elsewhere. One big trouble with trying to figure out the structure of the Earth is that you don't know exactly where the earthquakes are. You need to know the Earth's structure to compute earthquake locations, so it is a chicken-and-egg situation. I was talking about travel-time curves and so on, and I just glossed over how you know where an earthquake was and when it occurred. Of course to get the travel time, you need to know when it occurred, and to get the distance, you need to know where it was. Using an explosion, where you do know when and where it occurred, solves these problems brilliantly. They were, as I said, testing nuclear explosions in Nevada, which are equivalent to pretty big earthquakes. Some were as big as magnitude 6 or larger, and they generated beautiful, simple seismic waves.
An earthquake may actually rumble on at the source for several seconds, or even minutes for a big one, so you get very messy-looking seismograms. And if you are trying to unravel later arrivals caused by triplications in the travel-time curve, it is very difficult. You need sources that generate simple pulse-like waves. Nuclear explosions were a really good thing to use. They were of short duration and generated simple signals, you know where they were, and you knew when they occurred. And they were being well-recorded by the by special seismometer networks operated by the Air Force. Don had me looking at lots and lots of seismograms to try to figure out all these later arrivals having to do with structure in the upper mantle. And it became obvious, and it was obvious anyway, that the structure actually wasn't one-dimensional, it varied in all three dimensions. The geology at the surface corresponded to things underneath, too. Using nuclear explosions, since they were then pretty abundant, would be a good way to study the three-dimensional structure.
ZIERLER: How closely did you work with Don? Was he a hands-on mentor?
JULIAN: I'd say no. He put me onto looking at all these seismograms and measuring times of seismic-wave arrivals, and fairly quickly, he told me he wanted me to write a computer program to figure out travel times, so I had to learn Fortran. But mostly, no, I would go off, do stuff, and show it to him maybe once a week or something like that. We almost never sat down and said, "I've got this problem. The program doesn't work. What's the matter?"
ZIERLER: Did you do field work at all for your thesis?
JULIAN: Yeah. One of the data sources I used for my thesis was a nuclear explosion called Project Gasbuggy that was set off in northern New Mexico in December of 1967. Most nuclear tests had been conducted at the test site in Southern Nevada, but a few had been set off for various reasons in other places in Nevada, a couple in New Mexico, in the ocean off California, later some in Colorado, and there was even one in Mississippi. For Gasbuggy, I put out seismometers in a line extending west from southern Utah to Owens Valley in eastern California. Caltech had just, at that time, designed and built some portable seismometers. That was a radical thing at the time. Most seismometers were installed in permanent observatories. These "portable" instruments weren't very affordable, or very portable. A one-component seismometer was a trailer, which was powered, I believe, by eight heavy lead-acid car batteries and had film processing and developing equipment inside of it. It was really a big deal. Nowadays, a three-component seismometer can be about the size of a coffee cup, and you can purchase or borrow hundreds or thousands of them rather easily. But Jim Brune, in particular, had gotten Caltech to design and build a half-dozen or so of these trailers. I put them out in a line extending about along the 37th parallel, which is the border between Utah and Arizona then across southern Nevada and into eastern California. Collaborators from other institutions put similar equipment further to the east.
ZIERLER: What do you think your main conclusions or contributions were with your dissertation?
JULIAN: The main thing that Don had been trying to get at was the structure of what's called the "transition zone" in the mantle. The upper mantle was where everybody presumed a lot of geological processes must be going on. The lower mantle was imagined to be more passive and not doing very much. Then, there was the zone in between, but people didn't know what it was like, really. Gutenberg had published an Earth model that had the wave speed varying smoothly with depth, whereas Harold Jeffreys had published a model that had it varying smoothly with depth down to about 400 kilometers, then increasing more rapidly below that depth. Don had found that actually you could fit low-frequency surface-wave dispersion data (wave speed vs. frequency) better if you put in some pretty sharp jumps down at about 400 kilometers, where the wave speeds and density increased rapidly with depth. It wasn't clear whether there was one jump, or two, or something more complicated. He wanted to find out what the structure of this transition zone was in more detail, and surface waves, having long wavelengths, couldn't resolve it very well. The thing to do was to use shorter-wavelength body waves.
Basically, that was what motivated me to get me into this. The first part of my thesis is about the structure of the transition zone, even though the thesis ended up being titled Three-Dimensional Structure. In the end, I used data along three profiles from the Nevada test site, one profile from the Gasbuggy explosion in New Mexico, and several profiles from Project Early Rise, a series of large chemical explosions that had been detonated in Lake Superior specifically for seismological purposes. Seismometer profiles had been run from Lake Superior in about ten different directions across the United States and Canada, some of them nearly 3000 km long. That was a beautiful bunch of data, and it enabled us to figure out these triplications of the travel-time curve, and how they related to the structure of the transition zone. Now, here, you had this same kind of data in a whole lot of different places going across Canada, going across the United States, going across the Western United States from Nevada, so you could compare them. And it turned out, they were significantly different. Travel times at the time were thought to be measurable, if you pushed your luck, at a tenth of a second and maybe to vary by a second or something from place to place. But it became clear with nuclear explosion data from a lot of places, they varied by 10 seconds sometimes, which is a huge amount.
ZIERLER: I'm curious if you were thinking at all about plate tectonics or if it was a big deal at the Seismo Lab when you were there.
JULIAN: When I arrived, no. The conventional wisdom was that continents didn't drift. In fact, I talked many years later to a professor at Caltech, I won't mention any names here, who had himself been a student at Caltech and had wanted to write a term paper about the continental drift hypothesis and was told by his professor he wasn't allowed to do that, I'm sorry to say. [Laugh] I hasten to point out that this particular professor was an early one to change his mind on this question, and made many contributions, but people tended to have a closed mind about that question. And I remember when I first found out that that wasn't really right. There was a geology-club talk in the evening once a month or so, and at one of them Bill Menard, from Scripps, gave a talk about the fracture zones in the oceans. In about the last 10 minutes of his talk, he suddenly digressed and said, "I'm going to show you something brand new," and showed us these profiles of the magnetic stripes on the Reykjanes Ridge near Iceland in the Atlantic Ocean, which just showed beyond a shadow of a doubt that the ocean floor was spreading and making a tape recording of the reversals in the magnetic field.
There was no way around it. Things were moving horizontally on a very large scale. It was referred to at first as seafloor spreading because I think people imagined that the ocean floor was squishy or something. "Clearly, the Reykjanes Ridge is spreading, but maybe the continents still aren't drifting," and something in between was taking it up the motion. But Bill Menard's talk made it absolutely clear that the seafloor was spreading from the Mid-Atlantic Ridge, and by inference, probably from the East Pacific Rise and other similar features in the oceans. It had to be. That was in 1965. About three years later, it suddenly became obvious that the ocean floor wasn't squishy. It consisted of large plates that were almost perfectly rigid, and the concept of plate tectonics came along. Jason Morgan, at Princeton, and independently Dan McKenzie, who was visiting Caltech as a post-doc at the time, and Bob Parker at Scripps, realized that the Earth's surface was behaving like rather rigid plates, and you could use a simple mathematical theory due to Euler to describe the relative motion of points on different plates. That was in 1968, and the actual term plate tectonics came along a few years after that. For a while, it was called "the new global tectonics".
ZIERLER: After you defended, what opportunities were available to you?
JULIAN: Nafi Toksöz, who had been a graduate student and a postdoc at Caltech, had moved to MIT. He offered me a post-doc position there, which I'm extremely glad I took up. He was an enormously wonderful person to work with. There were other places I could've gone, but nothing could have been better than that. Frank Press was there by then, heading the department.
ZIERLER: What was the research plan for you when you got to MIT? Did you have a good idea of what you were going to work on?
JULIAN: No. They'd given me a post-doc paid for by ARPA to study "seismic discrimination" : distinguishing explosions from earthquakes. I didn't realize that at all. I sat down and talked to one of the people there and told them what I talked about, and they said, "What does that have to do with seismic discrimination?" I said, "What? I don't recall saying it did." [Laugh] But they had a reasonable attitude about the whole thing and realized they didn't know enough about seismology or the structure of the Earth, and that this was not a problem where you could look at some very narrow aspect of it and figure out how to do the problem. You needed to raise up the science, have better theories of wave propagation, a better understanding of the structure of the Earth, a better understanding of earthquakes themselves. I was encouraged just to study the structure of the Earth. While I was at MIT, I wrote papers about what's at the bottom of the mantle and things like that, which is obviously not where explosions are occurring.
ZIERLER: Was the appointment at MIT specifically in Lincoln Labs? Or was that an affiliation you had?
JULIAN: It was in the "Earth and Planetary Sciences" department. I was given a third of an office shared with two other people in the Green Building, which is gray, actually, a big tower where the geology department is, and another larger office, a couple of blocks away on Carlton Street, where the Lincoln Lab group was. Most of Lincoln Lab is 20 miles west of Boston, at Hanscom Air Force Base. But this one group, because they wanted to collaborate with the geology department, was on the campus.
ZIERLER: How long were you at MIT ultimately?
JULIAN: Five years. I left in 1975 and went to Colorado to work with the US Geological Survey.
ZIERLER: Tell me how that opportunity became available for you.
JULIAN: Dave Hill, who had been a grad student at Caltech at the same time I was, had been a Geological Survey employee in Menlo Park, California. They recognized that he was a very clever fellow, and the USGS paid for him to go to graduate school. One day in 1974, when he was back at the USGS in a senior position and I was at Lincoln Lab, Dave called me up and said, "Would you like a job working with the National Earthquake Information Center?" I said, "Perhaps," and ended up taking the job. The NEIC had, at that time, just been moved into the Geological Survey from the Coast and Geodetic Survey. Its job was to monitor earthquakes larger than about magnitude 4.5 occurring all over the Earth. They routinely published an earthquake catalog called Preliminary Determinations of Epicenters, which was one of the most important resources seismologists had. You'd try to find some seismograms for some particular study, and you needed to know what earthquakes had occurred. They were monitoring them, and publishing the information rather quickly. And that'd been taken into the Geological Survey and was moved to Golden, Colorado.
ZIERLER: Did you retain any of your nuclear-monitoring work when you arrived at the Geological Survey?
JULIAN: No, I didn't. Having just been taken from the Coast and Geodetic Survey into the Geological Survey, they were pretty disrupted. They had to do all their stuff on a new computer. They had it working on one kind of computer, and the Geological Survey said, "We don't have one. You'd have to use this one," which was enormously disruptive. The main thing I worked on there for the first year or two at least was just trying to get their programs working on this new machine, which was then yanked out from under them again. [Laugh] That's another story.
ZIERLER: In 1980, when you joined the Branch of Seismology, did you move? Was that also in Colorado?
JULIAN: I was in Colorado from 1975 to 1980, then I moved to Menlo Park in the summer of 1980. I don't think it was called the Branch of Seismology, but I couldn't swear to that because they changed the names of branches frequently.
ZIERLER: What were some of the ways that computers at this time were enhancing earthquake monitoring?
JULIAN: There are many, of course. Certainly, just enabling us to compute things like travel times or earthquake locations. When I first got a job at the Seismo Lab in 1963, people were using a globe and pieces of string to locate big earthquakes that had occurred. They'd say, "Here's the P-wave time and the S-wave time, so that tells us the distance. This earthquake is 5,000 kilometers from this station and 7,000 kilometers from that station," and they were using pieces of string to figure out where it had to be. Computers enabled that to all be automated and made much more accurate. We even eventually could take into account things like three-dimensional structure or the fact that the structure is complicated with depth and so on. Probably the first and most important one is just to compute things you couldn't otherwise compute.
The reason Don hired me in the first place was because he'd been studying surface-wave dispersion. Surface waves propagate along the surface of the Earth, but they feel down into the Earth. Shorter-period waves feel down into the Earth a short distance. Longer-period waves, which might have wavelengths of 1,000 kilometers or even longer, will feel downward hundreds of kilometers. And that's how you used them to learn something about the structure. You see how the wave speed is changing with the wave frequency. Computers enable you to figure out how surface-wave speeds in a depth-varying Earth will vary with frequency, which would be hopeless to do manually. You can try to do it analytically, and you're pretty well stuck to using flat-Earth models with one layer or at most, two layers, over a half space. With a computer, you can put in hundreds of layers, make it spherical, put in gravity, and do it right. But you couldn't possibility do that with slide rules or desk calculators.
ZIERLER: In what way did this work contribute to the overall networking of seismology monitoring at this point? Were computers connected with each other?
JULIAN: Oh, boy, no. Computer networking is a rather recent development. I mentioned a while ago that Stewart Smith at Caltech made one of the first digital seismographs. I've since learned that actually oil companies had done it earlier, but they were secretive about it at the time. But you can use a computer just to record and process data. That's the second important thing computers do. Besides letting you compute theoretical things about wave propagation that otherwise are impractical, they can actually be useful for recording data and also for doing various kinds of processing of it, filtering, in particular. If you record data digitally, you can get a computer to process them. The kinds of waves you're looking for can be emphasized, and various kinds of noise you don't care about can be suppressed. Digital signal processing is the second big thing that computers contribute to seismology.
ZIERLER: Tell me about your work in Iceland in the early 1990s.
JULIAN: When the earthquakes I mentioned at Mammoth Lakes occurred in 1980 and had mechanisms that people couldn't figure out and were causing people to do ridiculous things like throw out half the data in order to get an answer, I published a paper in Nature describing how I discovered, looking at these data, that you could perfectly well fit them. You didn't have to throw half the data out. They fit beautifully. But the thing that fit them in the theoretical source model was not rocks sliding past one another, it was a more general kind of source. It had a jargon-y name called Compensated Linear-Vector Dipole, and people had written papers pointing out that this was a possibility that didn't violate Newton's laws, but nobody had ever seen any. But these earthquakes from Mammoth Lakes looked exactly like the predictions for that kind of source model. I published a note in Nature saying that and saying I thought they were caused by dike intrusion or something like that. I then got a letter in the mail from Gillian Foulger, who I didn't know, who was a graduate student. She was both an Icelandic citizen and a Brit by birth, and she was a graduate student at the University of Durham, and she said, "I've got some funny earthquakes in Iceland I've been studying. I saw your paper. Maybe what you're seeing might explain what I'm seeing."
We met at an AGU meeting a few years later, where I got to look at some of her data, and we were indeed seeing something very similar. She had studied a geothermal area in Southwestern Iceland, about 20 miles from Reykjavik, that had a lot of these earthquakes, although much smaller than the largest Mammoth Lakes events. She came to visit the Geological Survey, and the Survey had a Gilbert Fellowship Program, to which people in the Survey could apply for money to do innovative things different from their regular job. They were named after Grove Karl Gilbert, who was probably the greatest American geologist of the 19th century and one of the greatest geologists who ever worked for the US Geological Survey. The first great one. Gillian and I applied for and received one of these Gilbert Fellowships to put out a dense digital seismic network in Iceland. She'd been using analog instruments, which are noisier and aren't as useful because you can't do the same kinds of processing. In 1991, we operated the network in the summer and early autumn for nearly four months and recorded thousands of earthquakes. And she'd, at that point, published a little paper in Nature saying that these earthquakes weren't double couples and wondering what they were. We definitely showed that was true. We got far better data with more stations and digital data, and the earthquake mechanisms were not double couples, period.
ZIERLER: In what ways did you stay involved in the three-dimensional structure area of geophysics?
JULIAN: More recently, I've been working a little bit on a method that developed probably in the 70s mostly called seismic tomography. If the structure's one-dimensional, varies just with depth, you can do what I described earlier, just measure the travel times as a function of epicentral distance, and you get a travel-time curve. It may be complicated, it may have upper branches and triplications, but still, it's just some kind of a multiple-value function of distance. If the Earth is three-dimensional, then everything is much more complicated. Travel time is not just a function of distance, it's a function of the three coordinates of where the earthquake was, three components of where the station is, and how the energy propagates along complicated rays that may be wandering around in three dimensions.
ZIERLER: When did digital seismographs come into the fore?
JULIAN: In the early 1960s, Stewart Smith built what was apparently the first digital seismograph in the academic world. It was located in a room at the Seismo Lab, on a relay rack, and it was about the size of a refrigerator. About half of this volume was a big tape recorder, using one-inch-wide magnetic tapes. It was just a one-off thing. It was recording data from some long-period Press-Ewing instruments located at the Seismo Lab, which were designed for recording distant earthquakes, not local ones. I think it had a sampling rate of one sample per second. If there was a big earthquake somewhere, you could go to Stew and ask him to get you seismograms from it, which were useful for things like measuring surface waves. But with one sample per second, it was not useful for travel times of local earthquakes or the sorts of studies I was doing, which used short-period body waves.
When I got to MIT, I found myself in an office with the Lincoln-Lab group that was studying seismic discrimination. The reason they existed, really, was to study data from the Large Aperture Seismic Array, which the military had just installed in Montana. It consisted of 600 digital seismometers deployed within a large, about 200-kilometer diameter, circle. After Stew Smith, I think that was the main next thing in digital seismology that happened, at least outside the oil industry. These were much higher-frequency instruments. They had a sampling rate of twenty samples per second, I think. A lot of what I did when I was at Lincoln Lab was looking at data from LASA because it told you lots of funny things about the Earth that you hadn't been able to tell with the more primitive instruments we had before. With this array, you could tell what direction the waves were coming from, for example, and you found that they weren't coming from exactly the great-circle direction, typically. There were all kinds of more complicated things going on because the Earth is three-dimensional.
ZIERLER: Was geothermal research something you did at the USGS?
JULIAN: I got started on that when Gillian Foulger contacted me about her non-double-couple earthquakes in this geothermal area in Iceland. This place where she had found these funny earthquakes is near a central volcano, and is where Reykjavik gets its hot water. We have worked together over the years, studying geothermal earthquakes, getting into this little business now of monitoring earthquakes for power companies because they're required to, and they don't know how.
ZIERLER: Tell me about the Mammoth 1997 Seismic Experiment.
JULIAN: I mentioned the 1980 earthquakes at Mammoth Lakes. They were followed by earthquake swarms that went on for years and were actually quite frightening. There were tens of thousands of earthquakes, accompanied by uplifts and subsidences of the ground of a sizable fraction of a meter, presumably caused by magma moving around in the volcano. In 1997, after this had been going on for 17 years or so, they had a really big swarm. By that time, portable digital instruments were far better (you could carry one in a backpack), and there was a pool of them owned by the National Science Foundation and managed by IRIS/PASSCAL, which lent them out for particular experiments. In fact, when we did the experiment in Iceland in 1991, we were one of the first users of this pool. In the spring of 1997, this swarm at Mammoth had been going on with increasing intensity for a few months, and we thought it would be a good idea to deploy a dense network to study it.
Jointly with Duke University, we applied for and got, from IRIS/PASSCAL, a bunch of instruments. Gillian Foulger, of Durham University in the UK, also applied to the British equivalent called NERC, which had a pool of the same kinds of instruments. Altogether, there were 65 or so three-component instruments. Then, lo and behold, about the time we put the first ones out in May, a bunch of intense earthquake swarms started occurring, and they got more and more intense throughout the summer. We were incredibly lucky. I mis-remembered earlier in thinking that we had done this in response to the swarms. It was the other way around. We were deploying instruments when the swarms started. This is almost unheard of. Usually, if you put a lot of instruments out, you stop all the earthquakes by Murphy's law or something. [Laugh] We were operating this extremely dense network of seismometers for about four months during this increasingly intense earthquake swarm.
Toward Early Warning
ZIERLER: Tell me about the development of the Early Bird Database.
JULIAN: The USGS runs a big network of several hundred seismometers in northern California, and in the 1980s it was not digital, it was analog. Each station converted its data–the voltage that comes out of a seismometer–to a frequency-modulated sine wave with a carrier frequency of about 80 Hertz. You then had FM versions of the seismograms. Those were then transmitted by radio to Menlo Park, where they were run through a discriminator, a frequency demodulator, which turned them back into the original voltage, more or less, and then that was digitized. They therefore needed to do something next, to notice earthquakes and save them, for example. Recording everything was beyond the capabilities of computers, or at least our budget, at the time. Rex Allen at the Geological Survey had constructed a special-purpose computer called the Real-Time Processor (RTP) to listen to these hundreds of digital signal and to notice when there seemed to have been an earthquake. Basically, it looked at the size of the signals going along, which most of the time were just noise caused by the weather and human activities, and noticed if it suddenly got much bigger, and thought, "Ah-ha, something has happened."
If that happened at several stations, it would then decide there had been an earthquake and save the data. It also did some extra processing to estimate the time of the earliest disturbance at each station, which is what you need in order to compute event locations. Rex's RTP was just sitting there, saying several times an hour, probably, "There was an earthquake here" and reporting a bunch of of P-wave timess. That's not what you need, though. You can't tell a newspaper reporter what the P-times were at the stations. You've got to tell them where the earthquake was and how big it was. That was the purpose of the Early Bird system. Someone had a put together a rather simple system, basically a shell script, a simple kind of computer program, which was taking the times and running them through a standard earthquake-location program. It didn't work very well, and it couldn't always keep up with the earthquakes. It was too slow. Early Bird was a system to try to improve all that. Again, it was still a pretty primitive program. Every now and then, the Real-Time Processor would say, "Here are a bunch of times for what I think is an earthquake." You would then try to locate it, figure out if it was an earthquake, and provide stuff you could give to the press, for example. "There was an earthquake 10 miles east of San Jose at 10:30 this morning, and it was magnitude 2," that kind of a thing.
ZIERLER: More generally, did Early Bird contribute to early warning in general?
JULIAN: It was a step in that direction. People were speculating about doing such a thing at the time, but now they have systems that detect the occurrence of an earthquake and notify people before the strongest seismic waves arrive, so that they can take protective action. When an earthquake occurs, nearby stations detect it quickly, and they can be used to warn people further away that, "Hey, strong shaking is going to occur." Early Bird did not do that. The computers weren't fast enough. That was the main reason. Basically, it was the same idea as the systems we have now. Early Bird would tell you that an earthquake occurred within about three minutes, but it couldn't tell you about it before the waves got to you.
ZIERLER: What was the mission of the Long Valley Observatory?
JULIAN: You probably should ask somebody else. It's more or less a bureaucratic concept. The Geological Survey has to monitor several volcanoes. Obviously, people live in Long Valley, especially during ski season, and Yellowstone is another obvious one. Mount St. Helens erupted in 1980 about the same time that all the earthquakes at Mammoth started happening, and there are many other dangerous volcanoes in the Cascade range. Alaska is full of volcanoes, which have disrupted airline traffic by making ash clouds in the atmosphere that airplanes have to avoid. There has been a Hawaiian Volcano Observatory since in the early twentieth century. Monitoring volcanoes is very important. You should ask somebody else about the exact history. They've basically just put names on these various aspects of monitoring volcanoes. Some of them correspond to something physical. There is an Alaska Volcano Observatory in Fairbanks in Alaska in a building, and the Hawaiian and Cascade observatories also have physical manifestations. The Long Valley Observatory is more of a conceptual thing: people in Menlo Park continuing to monitor earthquakes at Long Valley. It isn't a new building or anything like that. The Yellowstone observatory is also not an actual building.
ZIERLER: Tell me about doing algorithm work to study earthquake mechanisms in a more precise way.
JULIAN: As I was saying, these earthquakes in 1980 at Mammoth and the ones Gillian Foulger had observed in Iceland had mechanisms that were more general than the "double couple" force systems that correspond to rocks sliding past one another, the process that had been the default assumption of seismologists before that.
The way you figure out a double couple earthquake mechanism–there are many ways, but the simplest way is, you look at the direction of the first wiggle on the seismogram and whether it goes up or down. If it goes up, that means that motion was outward, away from the earthquake, because of the way the waves travel through the Earth. If you work your way back to the earthquake source, the first motion from the earthquake might be away from the source, or it might be toward the source. For double couples, it's evenly divided. On one half of the focal sphere–a little imaginary sphere around the earthquake–the first motion will be outward, and on the other half it will be inward. And these parts of the focal sphere will be divided by two orthogonal planes. If you think of this focal sphere as being the globe, you could imagine the equator as one plane and the prime meridian and the date line as the other.
These planes divide the sphere into four quadrants. In two of the quadrants, the motions go outward, and in two, they go inward. To figure out the mechanism, plot the first motions on a map of the focal sphere and try to find two orthogonal planes that divide them. There are standard graphical methods for doing it. But when you get to non-double-couple earthquakes, they aren't planes anymore. In fact, the motions could all be outwards. An explosion is one kind of non-double-couple mechanism. You can't use the same graphical methods. So we figured out ways of using a computer to do it. The method you're probably referring to here uses linear programming, a mathematical method that was originally developed for economists. Linear programming can deal with inequalities, and a first motion is basically an inequality, saying that the P-wave amplitude is either positive or negative.
ZIERLER: What were some of the collaborations or projects that led to your honorary affiliation with Durham University in the UK?
JULIAN: Gillian and I have been working together since our earthquakes brought us into contact with each other. As a result, she visited the Geological Survey on many occasions. She actually came and had a job for a year and a half at the Survey, and I've gone to Durham and worked there with her. I actually don't remember how I came to be an honorary research fellow. I think they wanted it. We were publishing papers, and if I could put "honorary research fellow, Durham University" on it, then they got more brownie points somewhere. [Laugh]
ZIERLER: When it was time to retire, was the plan to do geothermal consulting? Was that already solidified at that point, or that came later on?
JULIAN: We were already doing that, yeah. For the last twenty years or so before I retired, we had done studies looking at earthquakes in quite a few different geothermal areas. This started shortly before we went to Iceland, in the summer of 1991. These were, as I said, brand-new instruments that the IRIS/PASSCAL pool had, and we'd never used them before (nobody had), so we thought it would be a good idea to make sure we knew how they worked. We arranged it to do a one-month study in April of 1991 at The Geysers geothermal area. That's the largest exploited geothermal area in the world, actually. It's north of San Francisco about two hour's drive, and provides 5 or 10% of the electrical energy in Northern California. And for most of the last few decades, it's been the most seismically active area in California. It was only slightly seismically active before 1980, but when they started generating power there, it started having earthquakes. And the earthquakes have increased with the power production over the years. That's how we got into studying geothermal areas, besides the fact that Gillian's earthquakes in Iceland were in a geothermal area.
ZIERLER: Moving the conversation even closer to the present, tell me how you got involved in looking at seismology on the planet Mars.
JULIAN: All this time, we've really been talking about volcanic earthquakes, because volcano and geothermal are almost interchangeable words. One of the most interesting volcanic phenomena is volcanic tremor. It's tremor, not tremors. For a long time, whenever I gave a talk about this subject, someone would change my title and put an S onto it as though I must be talking about earthquakes. But it's not earthquakes, it's a continuous vibration of the ground that is very often observed around active volcanoes. It's essentially always observed when they're erupting, and it's almost always accompanied by evidence, such as ground deformation, that magma is moving around underground. It's a continuous vibration of the ground that can be quite strong. I've been told that people who lived near Kīlauea often were awakened at night by volcanic tremor, the shaking was so strong. In fact, they said during a big eruption, they would awake in the night when it stopped because they had gotten so used to it. [Laugh] Anyway, people had never really known what caused this, except it was clearly associated with magma and possibly other fluids like steam or water underground.
Somewhere, I got the idea that it operated by a mechanism similar to the one that makes your vocal cords work. Basically, what happens is, if you've got fluid flowing through some kind of a channel, and the walls can move, when the fluid speeds up, the pressure drops. That's called the Bernoulli principle, and it's slightly mysterious. That would cause the walls to move inward, which would restrict the flow, slow it down, and make the pressure go back up. Fluid flowing through a channel can make it oscillate. Things like this are everywhere. Why should a flag flap in the wind? If you blow on it, why doesn't it just move over a bit? Why does it go back and forth? I investigated this mathematically and came up with an equation that I thought might describe tremor. It's a very messy nonlinear differential equation, and when you put it into a computer and solve it numerically, it does produce oscillations that look very much like real volcanic tremor. And it does things that hadn't been seen in tremor, but that have been noticed subsequently.
ZIERLER: What have been some of the most surprising things we've learned about seismology on Mars?
JULIAN: I haven't gotten to Mars yet, I was just talking about tremor. That's how I got into Mars. A friend of mine, Sharon Kedar, who works at JPL now, was interested in tremor, too, and he was at the Seismo Lab. I believe he contacted me and told me that the Insight seismometer on Mars was seeing signals that looked like long-period earthquakes, which are a phenomenon very similar to volcanic tremor. They weren't sure what they were, and with one seismometer, they weren't even sure where they were, but they had a good candidate, this place called Cerberus Fossae, which has big fresh-looking cracks in the ground. They are very spectacular. They're clearly very young fissures. Cerberus was the dog that guarded the gates of hell, in Greek mythology. The people at JPL were seeing something that might be volcanic tremor or long-period earthquakes on Mars coming from the direction of these apparent volcanic fissures called Cerberus Fossae, and Sharon wanted to apply this theory I had developed to these observations. I worked with them a little bit on that, but they did most of the work, which was trying to figure out which parameter valuess in my model would give predictions most like the seismograms they were seeing on Mars. This is an ongoing story. With one seismometer, we don't know much. I hope we get more seismometers. So in answer to your question, I think the most surprising thing we have found about seismology on Mars is these possible volcanic earthquakes.
ZIERLER: What are some future missions that might include more seismometers?
JULIAN: You'd need to ask somebody at JPL about that.
ZIERLER: But is your sense that there's ongoing interest in this kind of research?
JULIAN: I certainly hope so. Doing seismology with one seismometer is extremely challenging.
ZIERLER: Just to bring our conversation right up to the present, what are you currently working on?
JULIAN: I started to mention seismic tomography. Seismic tomography uses travel times observed at a whole lot of stations, preferably from a whole lot of earthquakes, to figure out the structure in three dimensions. It can be done various ways. You can do it with local earthquakes, which we did at Mammoth in 1997. You can do it with teleseisms, where you just put some seismometers out and wait for earthquakes to happen all over the world. And the waves coming up under your seismometers will tell you something about the structure under your seismometers. A lot of people have been using tomography to do all kinds of things, but quite a lot of them have been claiming to see temporal changes in the wave speeds. It's very likely that wave speeds do vary with time. In fact, they certainly do. Some variations can be measured for sure. Some wave speeds vary with the tidal stresses in the earth. However, many of these claims of temporal variation are suspicious. Measuring temporal variations with tomography is challenging because the earthquakes aren't all in the same place.
The structure's three-dimensional, and you have an earthquake, and the seismic waves sample along a bunch of rays to some stations, then another earthquake happens in a different place and samples a bunch of other paths, so you get different wave speeds out of it. You might say, "Oh, the wave speed changed." That's very dangerous. If you measure anything twice, you'll get different answers, because all measurements are inaccurate. In this case, changes in the locations of the earthquake are causing you to sample different areas. That's not an error, but it's a source of noise if you're trying to look for real changes. Basically, the solution to this problem is, when you're doing the tomography, don't just solve for the wave speeds. You should try to solve for all this stuff at once and find out what the possible ambiguities are. This is still something I want to work on. I haven't been working on it enough.
ZIERLER: For the last part of our talk, I'd like to go back and ask a few retrospective questions. Given your time at the Seismo Lab, what did you learn about approaching the science, the collaborations, the big questions in geophysics and seismology, that has stayed with you throughout your career?
JULIAN: That's a good question. I sometimes was accused at the Geological Survey of being too much of a Lone Ranger and just working on things I was interested in, rather than the things I was supposed to be working on. Quite honestly, I think if people work on the things they're interested in, it'll be more productive than if they're being told, "We have a big project to do such-and-such. You've got to solve this problem, even though it's not something you've been thinking about or are very interested in." I think that would be what I've learned. I suppose people who run research institutions don't like this, but I think people will be most productive when they're working on things they're curious about and interested in.
ZIERLER: When has that served you best, when you pursued research that was most interesting, even if it meant you had to go at it alone?
JULIAN: Always. Everything I've done that people ended up thinking was great and wonderful, I was nervous that someone would look over my shoulder and see what I was doing, and it wasn't what I was supposed to be doing. [Laugh]
ZIERLER: Finally, last question, looking to the future. Do you think that approach is still advisable for younger scientists in the field?
JULIAN: I don't know. Clearly, there's a trend for the last hundred years, probably longer than that, for there to be more large-scale cooperative projects, monster projects, papers about gravity waves that have a thousand authors. Obviously these projects can be very productive. They have done something I would've thought was impossible. But that's a profound question I don't know the answer to. Some sort of a balance between people doing what they're curious about and that motivates them and things that require massive efforts and many, many kinds of specialist knowledge
ZIERLER: On that note, I'm so glad we were able to do this and capture your recollections. Thank you so much.
JULIAN: Thanks a lot.