skip to main content
Home  /  Interviews  /  Timothy Melbourne

Timothy Melbourne

Timothy Melbourne

Professor of Geophysics, Central Washington University

By David Zierler, Director of the Caltech Heritage Project

September 2, 2022

DAVID ZIERLER: This is David Zierler, Director of the Caltech Heritage Project. It's Friday, September 2, 2022. I'm very happy to be here with Professor Timothy I. Melbourne. Tim, it's great to be with you. Thank you for joining me today.


ZIERLER: To start, would you please tell me your title and institutional affiliation?

MELBOURNE: Yes, I'm a Professor of Geophysics at Central Washington University. I went there straight from Caltech, and I've never left. [Laugh]

ZIERLER: Tell me about geophysics, seismology, geology at CWU. How long has the program been in existence?

MELBOURNE: Geological science as a major, I think, has been around for a very long time, probably 1920s to 1930s. The way I ended up there is directly attributable to my experience at Caltech. I got my PhD right around the time GPS geodesy was getting quite mature and had kind of reached the level where it could be done in places outside of federally funded research centers, where you have 100 PhDs in one building, like JPL. Using GPS, I studied the 1995 Jalisco earthquake, and I knew what went into building a GPS network, what went into the data analysis, what went into the models. And it always struck me as a graduate student that there was no continuously operating geodetic network in the only subduction zone in the lower 48 that mattered, which was Cascadia.

Towards the end of my graduate career, I started to learn about a group at Central Washington University that had started to build some stations out. I defended my thesis in 1999, and it was obvious that if there was going to be a GPS network in the Pacific Northwest in the Cascadia Subduction Zone, it was going to be run out of Central Washington University. And it was a nice place to live, to boot. It was therefore kind of a very easy jump for me. And it turns out, that was a way to stay close to the data and the science, and it's worked out very well. I've been here 23 years, and it's been one new thing after another. That's why I came here. At the time, the Cascadia Subduction Zone had five operating stations in 1999, and now, the network that CWU operates, PANGA, the Pacific Northwest Geodetic Array, has 270 stations ranging from Cape Mendocino, which is the southern end of the Subduction Zone, up to the Canadian border. And we exchange data with the Canadians.

The decision to go from Caltech to Central Washington University was totally a function of how Caltech molded me. CWU was a good decision for me on many levels, although it did raise some eyebrows at the time because CWU does not have a PhD program. More than one faculty said, "Are you going to be able to get any research done between your teaching loads?" I said, "I don't know, we'll see." CWU is not a research university, it's not that kind of place. On the other hand, the guiding principles and values, and training, and intellectual outlook on life that Caltech really gave me has, perhaps to the consternation of my colleagues here, constantly redirected me back towards research.

ZIERLER: When you talk about the maturation of GPS and seismology at the turn of the century, is that more technological or administrative?

MELBOURNE: Very much technological. As software developed, algorithms improved, bugs got fixed, computers got faster, things got easier and you could do things on a desktop computer that required, 10 years before, a lot more infrastructure, both people and machines.

ZIERLER: As the only one in the lower 48, I wonder if you could speak to what's so unique about the Cascadia Subduction Zone.

MELBOURNE: Well, from a seismological, geological, or geophysical standpoint, there's really nothing unique about it other than that it's ours. It's under Washington, Oregon, Northern California, and British Columbia. In terms of rates of convergence, it's actually fairly slow, quite a low convergence rate, just under 4cm/year. That means you don't tend to see as much activity as you do when you go to very, very active places like Japan or Costa Rica, for instance, where if you stay there for any extended period of time, you're going to feel an earthquake. That's not the case in Cascadia. It's also a beautiful part of the world, that was important to me as well. Then, in time, the geodesy showed that the Subduction Zone is actually extremely active, the fault itsetlf. You've just got to know how to look at it to see the activity. It has this phenomenon called slow-slip earthquakes, and for the first six or seven years following their discovery they were only detectable through the crustal deformation measurements that GPS provided.

In time, seismologists in Japan, using a very quiet network of borehole instruments, were able to tease out a seismic signature to these slow-slip events. Once they showed the world what the seismology looks like during one, everybody else could tune filters and build arrays, and now, we see these events routinely. It's a very esoteric seismic signal, just a form of noise, really, and a high-frequency signal. Much of the initial work on this phenomenon came out of the GPS crustal deformation measurements in Cascadia. And it's a gift that keeps on giving, because one of the things we figured out was that unlike most earthquakes, which are unpredictable, in one part of Washington state and Southern BC, these slow-slip events have a drum-beat regularity every 14 months. So you could assign network deployments around their anticipated date and be ready to capture it, which is really a remarkable thing in seismology. We published that paper in 2001, in Science, and that drum-beat periodicity we observed then is still going on today, 22 years later. But that's a very special part of the Cascadia Subduction Zone, other regions have other periodicities or none at all. So this is just one aspect of Cascadia that's very interesting.

Another interesting aspect is that the earthquake culture in Cascadia, meaning Seattle, Portland, Everett, Olympia, Eugene, is not nearly as developed as it is in California. When you combine a lack of earthquake culture, which is manifested today in nowhere near Japan-level building codes along with a historical legacy of dangerous buildings that wouldn't be built today, unreinforced masonry and the like, and then you combine that with a very big, nearby subduction zone fault and you do have the makings for a substantial catastrophe. The odds that we see it in our lifetime are pretty small, however. It's been 320-some years since the last one, and we know the recurrence interval. But when I'm teaching about the slow slip events in my classes, I use the analogy that if you've got your head in the guillotine, and the blade up there creaks a little bit, that's interesting. [Laugh] That's a little bit what these slow-slip events are like.

ZIERLER: I wonder if you can talk a little bit about the concept of extrapolation. In other words, your close proximity to Cascadia Subduction Zone and the fact that, besides it being "ours," it's not terribly unique. With all of that in mind, what have you learned about it that says things more broadly on a planetary scale about subduction zones?

MELBOURNE: Even as mellow or unexcitable a subduction zone as Cascadia, if you have the tools and know how to measure it precisely enough, it's extremely active, very much alive. We can see slow slip events happening again and again, propagating along the fault, interacting with each other. It stands to reason that all these other convergent margins that don't have the instrumentation density that Cascadia has are also probably very active at low levels. Certain places, most notably Japan, which has the world's best geophysical networks has certainly shown that to be true, we know that. They've demonstrated incredible levels of activity in between what normally be called earthquakes, far below what most conventional instrumentation can see. But at least in the northern part of Japan, where the convergence is about nine centimeters a year, whereas Cascadia is less than half that, but we still see all kinds of activity. This tells me that if you can put a lot of instrumentation in, you can learn a lot about the details of how these things evolve with time.

ZIERLER: What you've found about periodicity, does that get us closer to understanding earthquakes as cyclical events?

MELBOURNE: I wish that were the case, but I'd have to say the preponderance of evidence is no. If we could watch this periodicity for a couple hundred years we might get a hint of changes related to an incipient rupture. There are many models and fault-dynamic models that suggest the periodicity we observe should change over the course of the earthquake cycle, meaning hundreds of years, as a big fault matures. This is because the accrued stresses imparted by hundreds of slow slip events on the locked portion of the fault add up, and by so doing push it towards the rupture that we would call a big earthquake. There are also reasons to think the locus of where this slow slip happens should start to propagate updip, towards what we call the seismogenic zone that yields large ruptures. So if we could observe Cascadia for a couple centuries and see such a trend, then the answer would be incredibly instructive. But we've only been doing this for 25 years, and that's a very small delta-T (time interval) when you're talking about a recurrence interval of 5-600 years.

ZIERLER: I imagine if it's not useful yet for understanding earthquakes as cyclical events, that won't tell us very much that could get us closer to earthquake early warning or, dare I say, earthquake prediction.

MELBOURNE: Yes, the reason I'm hedging here is that earthquake prediction is a magnet for really bad science. There must be 100 years of really bad science based on attaining the holy grail of predicting earthquakes. When you start to get into the idea that certain things are periodic, that leads very nicely into the idea, "If it's periodic, and it's been some amount of time since it happened, you can predict when the next one is going to happen." And yes, we do that with these slow-slip events, but only these slow-slip events, and only one part of Cascadia under Puget Sound. But one of the things we've really come to understand, both with Cascadia and with other subduction zones around the world, is that there are certain questions in science that are essentially unanswerable and almost not the domain of science to even attempt to answer. For instance, what's the weather going to be at 2 pm in Pasadena on June 19, of the year 2351? There are so many nonlinear dynamics, many of which we don't understand but will control, to first order, the answer in there that it's not really a scientific question to ask. If you go to the earthquake fault world, what we find, and what instrumentation in Cascadia and elsewhere, is that there's not just a big earthquake, then 500 years of nothing. What we see instead is a very nonlinear, complicated world where you have slow-slip events that can trigger other slow-slip events, which can in turn trigger real earthquakes that people feel, and then that seismic slip triggers other slow-slip events, which then propagate down dip, or up dip, or sideways, which then trigger other earthquakes. There's a feedback going on between slow and fast slip occurring on many faults and happening on a continuous basis. So to say, just because you're seeing a 200-kilometer north-south distance of a 1,500-kilometer subduction zone exhibit a periodicity of 14 months, between, say, 1988 through 2022, that it means anything for the greater subduction zone is just not supported by the observations we have. It's an extrapolation unsupported by any observation or theory we can defend with data.

So we've come to the conclusion that in the absence of any paradigm-shattering discovery, the predictability of earthquakes is nonexistent. And the combined seismic and geodetic instrumentation we've studied in Cascadia shows this to be absolutely true. There are a couple beautiful examples of such unpredictability made in the Ecuador subduction zone by a French group. It's a very active subduction zone, and they have had many big earthquakes. Recently they had a small earthquake, which triggered a slow-slip event such as we see in Cascadia, which then triggered another slow-slip event, which then triggered a big earthquake, which then triggered another slow-slip event, which propagated back up to the original location of the first slow slip event and kicked off a big aftershock right there, all over the space of a couple years. To be able to use that observation to make some measurement of predictability is nuts.

ZIERLER: As you've already alluded to the idea of a nonscientific question with all of the nonlinear dynamics at play, it's a question probably as much philosophical as scientific. If the prevailing idea nowadays is that earthquakes are unpredictable, it seems to me it still leaves us with two options. One, that's a statement of our limitations, both in observation and theory, or it suggests that earthquakes are fundamentally chaotic, that there's no scenario in the future where, no matter our technology, theory, or observational skills, there's nothing fundamental to earthquakes that could tell us when they might occur. I wonder what your thoughts are.

MELBOURNE: I think it's the latter. Again, barring some earth-shattering discovery, there's nothing I've seen that survives really good scrutiny, anything precursory, that you could actually point to. Going back to the weather analogy, to be able to exactly pin down what the temperature in Pasadena at 11 am on June 19 in the year 2351 might be impossible, but you can put some statistical bounds on it to the extent that you can make measurements, understand what the climate is likely to be in Pasadena roughly at that date, and put some statistical bounds on the range of temperatures you might expect on that date in the future. In the seismic case, such bounds are useful because you can build appropriately. If you were coming in from another planet and setting up shop on earth for the first time, you would want to know this information as you built your cities. I get this question sometimes from press here in Washington State about, "If earthquake predictability is impossible, what good are you?" essentially. These people are not trained as scientists, so it's a perfectly reasonable question for them to ask if they're not aware of the complexity issues. So there's an analogy I give. "Your incipient car accident isn't predictable either, but you go through your life making some plans. You put on a seatbelt, you buy a car with airbags, or you vote for the politician who will mandate that airbags be included." When I'm talking to the press in the Northwest, I'll say things like that. "If you know earthquakes are coming, eventually, and you can characterize how big and long the shaking may be, where the tsunami inundation is likely to be greatest, you can make some decisions today to protect yourself or your grandkids in the future. It's kind of the equivalent societal decisions as making people put on seatbelts to protect themselves for a future risk whose probability is very much nonzero. Decisions like don't build your house in a tsunami inundation zone, don't allow unreinforced masonry to be built anymore in Seattle, that sort of thing. All of this adds some value to the measurements we make and the work we do.

ZIERLER: A parameters question, what are the thresholds by which you determine a slow earthquake? How slow is slow, and when do you get to fast?

MELBOURNE: That's a good question because it turns out, the more broadband the measurements you have to study these things, the more you realize the rate of slip and duration of these events actually fills much, of not most, of the observable slip spectrum. But we know, simply from an instrumentation standpoint, that we have big gaps in our detectability window. But in Cascadia, when these slow slip events turn on, we typically see about one cm of surface deformation that indicates a couple centimeters of slip along the fault, typically between 25 and 40 kilometers' depth, which is a lot deeper than the seismicity of Southern California. That couple of centimeters of slip accrues over, say, four to five days in any given spot and the slow slip locus migrates along the subduction zone fault for typically 3-4 weeks. And the total moment magnitude of the resulting slip is usually in the mid- to high magnitude 6 range in Cascadia. Compare that to a conventional earthquake, like the 1994 Northridge M6.7 earthquake, in which maybe 10 to 15 centimeters of slip over a roughly 10-kilometer patch happened in five-to -six seconds. When you spread the slip out over days, it doesn't produce the shaking, which is why it wasn't ever identified on seismic monitoring. But that fault slip still deforms the earth, which is why the GPS geodesy could detect it originally.

ZIERLER: We talked about the maturation of GPS at the turn of the century. Where is it now? What are some of the key technological advances 20-something years later?

MELBOURNE: This is really a close-to-home question because when I got to CWU in 1999, we had five or six stations, and we had modems attached to these stations. The GPS receivers would stream to a data-logger. None of my students know what a modem is anymore. Because the phone rates were cheaper at 2AM, some chron would kick in and dial up some stations over a 56k modem–that was actually toward the end of the modem world–and download this data. In time, starting around maybe 2002, 2003, GPS receiver manufacturers started putting ethernet chips on board the receivers. It was just cheaper to use ethernet, for them and for us. That got rid of modems, and this is really interesting because now, with our GPS receivers streaming data continuously onto the internet, we had to deal with a new problem of catch the data continuously at CWU as it's being acquired. And because the speed of light is fast, whether the data was coming from Seattle, or the coast of Washington, or the coast of Oregon, or the coast of British Columbia, it shows up at CWU in a tenth of a second. So we kind of had to rethink how and what we were doing, and why. We were no longer using modems to download batches of data every 24 hours, but streaming it in continuously with essentially no delay.

So it took no thinking whatsoever to realize that if we're getting data from all over Cascadia, effectively instantly, we've got the makings of a seismic network! One that isn't built out of conventional seismometers, where you've got a magnet inside a coil with some feedback circuitry buried in the earth. Instead, you've got a GPS antenna firmly anchored to the ground but listening to overhead satellites, with data streaming in real time. And furthermore, since we know how earthquakes work if we actually get a really big one, some of these GPS stations are going to move five meters, say, and we'll see that motion almost instantaneously. So in 2006, I wrote the first proposal, to NASA saying, basically, "Look, this is bleeding-edge, pie-in-the-sky stuff, but we've got 50 stations out along the Cascadia Subduction Zone whose data we get in a tenth of a second, day in and day out, and we can build a very coarse seismic network out of this." And we had to be really careful because a good seismometer, the Pasadena Streckeisen say, has a sensitivity to velocity roughly a million times finer than what GPS can do by differentiating position through time. On the other hand, if you've got an earthquake that's moving your station's five meters in half a minute, you actually don't want a super-sensitive instrument for that, you want a blunt one, for all sorts of reasons. So I pitched to NASA, "We think we can build a very blunt GPS-based seismic networks that will tell you, if you have good coverage, the magnitude of an earthquake within the time scale of the fault rupture itself." For a big earthquake, mid-7s say, it might be 30 seconds to a minute during which the fault itself is actually sliding. So we basically said we can capture that information instantly and tell you nearly instantly how big your earthquake was. And NASA bit.

That started the project I've been kind of working on for the better part of 12 years, building a strong-ground-motion, meaning blunt, GPS-based seismic network. We started in Cascadia, and the rest of the world caught up in terms of building out networks, so we've now expanded it globally. We're pulling in a couple thousand stations from around the world. Data from Antarctic GPS stations come in within half a second, so does New Zealand's, Chile's, Oceana, Indonesia, etc. We position these data within a very well-defined reference frame, and then we take those positions and send them on to the US agencies that are responsible, and required by law to be responsible, for certain hazards. For instance, USGS handles earthquakes, NOAA handles tsunamis, and NASA technically isn't supposed to handle either, officially, but sometimes they still do just by virtue of being on the bleeding edge of technology development. We send the GPS positions we estimate from stations all over the world to the USE National Earthquake Information Center in Golden, Colorado. For the West Coast ShakeAlert earthquake early warning program, we send them to Caltech, Berkeley, and the University of Washington. We also send subduction zone-relevant GPS position streams to NOAA's tsunami warning centers, and they use them with a variety of different algorithms to rapidly characterize earthquake magnitude, since earthquakes have caused about 80% of tsunamis over the last century. They're interested because if the wait on the Global Seismic Network estimation of magnitude it's about a 15 min delay. For instance, if you have a big earthquake, say, in Tonga or the Solomon Islands, and you wait for the P- and S-waves to propagate out through the planet, and use what is commonly called the W-phase inversion (which came out of Caltech, Kanamori and Anderson, early 90s), which is the standard way of characterizing a distant big earthquake quickly. But that 15-min delay is greater than the tsunami inundation time at the local coastline. So if you have acess to GPS in sufficient coverage to characterize the earthquake, you can get a good magnitude estimate much more quickly. In practice, for tsunami warning, the agencies are very conservative, and they have to be. If they have what smells like a big earthquake recorded on one local high-frequency instruments, they hit the red warning button, and if there's no tsunami they can cancel later. But we've now had a couple examples, like up in Alaska, where there was no geodetic offset, which meant the focal mechanism of the earthquake is probably not tsunamigenic, and this allowed them to cancel the alert very quickly. That helps them because people get inured to alerts if they're constantly coming but then there's no tsunami. It helps them avoid that problem, helps build confidence in the system, helps them have confidence in their solutions. So including GPS into the tsunami warning program came naturally out of this evolution of the technology, the same that allowed our network, PANGA, the Pacific Northwest Geodetic Array, to march on the heels of the technology and surf that wave, so to speak.

ZIERLER: Are you a founding member of PANGA, or does it predate your tenure?

MELBOURNE: It actually predates my tenure. The two people who hired me were a husband and wife team, and the wife, Meghan Miller, I think actually wrote the first proposal to put the first station in the ground. She was a geologist who post-doc'd at JPL and moved up here about four years before I did. Her husband, I think, was the one who actually came up with the name PANGA. It's a pun, because they both liked Mexico, and a panga is a Spanish word for a Mexican fishing boat. I arrived at CWU in 1999, and by 2003 or so, Meghan had decided to go into administration and became a dean. So the lab and running the network was unceremoniously dumped on me, [Laugh], but I had other things to do. I had Don Helmberger yelling at me to get my thesis published because his funding was running into trouble [Laugh] so I actually had no real interest in running it, but somebody had to do it, and today I'm very glad I stepped up.

ZIERLER: A few overall questions to your research. Among the fields and disciplines, there are seismology, geophysics, geology, earth science. At the end of the day, what would you call yourself? What's your home discipline?

MELBOURNE: Geophysics. I went to graduate school to do seismology, pure seismology, and this is–is Hiroo going to listen to this? [Laugh]

ZIERLER: I hope so. [Laugh]

MELBOURNE: All right, Hiroo, this one's for you. I went to study seismology, and that's what I intended to do. However, between undergrad and grad school, I worked for the USGS for two years at the Cascades Volcano Observatory, and they were putting GPS on volcanoes. That's actually where I first learned how to build a GPS station, and did so on some very fun volcanoes, like St. Helens, Rainier, one called Augustine up in the Cook Inlet of Alaska. At Caltech, you had these thing called oral exams. In theory, they want to make sure you know what science is, and they kick you out if you don't do well enough. You have to have these two independent projects. One of mine was to work with Joann Stock, who was building a geodetic network in Jalisco, Mexico, which is also a subduction zone. She was interested in the continental dynamics aspect because it's a big uplifted terrain that hasn't quite busted off mainland Mexico and become attached to the Pacific Plate the way the Baja Peninsula has.

If you look on a map, Baja is here, then Jalisco is right here under Baja. And Jalisco is undergoing that detachment process right now. In 10 million years, it's going to be attached to the Pacific plate following Baja up into the Aleutian Trench. One of my oral projects was working with Joann on this data. But the continental deformation rates in Jalisco are very low, indeed, and I was like, "I don't want to be a grad student for the 12 years it's going to take to get a resolvable signal here." So I went and talked to Joann, and she was incredibly gracious. She was like, "Yes, it'd be a while. You came here to do seismology. Go thrive." And that's one of the really special things about the Seismo Lab, and I hope it's still like that. Students can just follow their interests and the vicissitudes of fate that throw projects at them. So anyways, maybe a year later, I think I was auditing a class Hiroo was teaching, and Hiroo came in to this class one day and said, "Tim, there's been a very large earthquake in Jalisco under your GPS network." My head space at that time was in trying to understand Cagniard-De Hoop, which was a totally esoteric way of solving a wave equation, and other stuff Helmberger and the courses there had thrown at me, I wasn't thinking about earthquakes. I looked at Hiroo and simply said, "Huh."

Hiroo looked at me and said, "Huh?" [Laugh] I couldn't process it. I was, essentially, "What does this mean? What should I do?" In time, it turned out it was a really big earthquake. Knowing what I know now, I know that that event was, basically, 10 PhD theses right there, but at the time I couldn't process it. And I remember Hiroo just looked at me, seeing some kid who was dumber than a box of rocks and couldn't seem to process that there'd been a massive earthquake under their geodetic network. Of course, Hiroo knew everything that was going to come out of it, but I didn't, I couldn't as a early grad student. But wow, talk about fate handing me a golden egg. I was watching industrial strength Caltech seismology ongoing about this subduction-zone earthquake under Jalisco, Mexico, where Joann, I, and her crew had built this network and gotten a measurement. That's the key thing, they installed it, built it, and got the initial measurement, so we knew where stuff was at the centimeter level before the earthquake.

Anyway, it took me a day or whatever, but I very sheepishly walked back into Joann's office and I said, "Hey, if you don't have any students to work on this, I'm not that stupid. I'll work on it." And she was very gracious, as always, and said sure. We saw all kinds of stuff, interesting processes. It was October of 1995 when the subduction zone had broken in a magnitude 8 earthquake, but what we found it that the fault wasn't done sliding in the weeks following the mainshock. It, too, kicked off all these slow-slip events, creep events which were clearly triggering aftershocks. But we could see this only because we had flown down there and put all this GPS out before the earthquake. It took us four or five days to get down there after the earthquake. So I ended up working on that earthquake for a couple years and published a whole bunch of interesting papers. Then, there were other faculty at other universities with other grad students also involved with the project, so eventually I had to move onto other stuff and share the wealth, so to speak. In any case, that ended up leading to an understanding of how to put in GPS, analyze it, monitor it, and all that, which would prove very useful in my move to Cascadia in subsequent years.

ZIERLER: What aspects of your research agenda are more on the theoretical side, and which are more on the observational/experimental side?

MELBOURNE: The vast majority of my time over the last 10 years has been spent with my group trying to increase the precision and reliability with which we can position a receiver to centimeter accuracy. In conventional GPS positioning, you download 24 hours of data, so you have a ton of measurements, from which you estimate a single position based on 24 hours of continuous measurements. What we do now is to make a position estimate every second, like a one-sample-per-second seismic instrument, but you have one second of data to update a position. So you end up having a much larger scatter and lots of other vicissitudes creep into this kind of undertaking. So most of my effort has been working on trying to improve the reliability and resolution of GPS positioning in real-time. We like to use legumes as a metric. If we could measure the phase center of an antenna to the volume of a lentil, we could really see fairly small earthquakes, low magnitude fives. But we're somewhere between an orange and a grapefruit at the present time, continuously 24 hours a day, which places an effective floor of about mid-magnitude 6 in what we can image. . What we do is not dissimilar to seismology. GPS measurements are, at the core, a ranging measurement, meaning if you know where your satellites are and can get a range to the satellite, then you have a sphere on which the position must lie and if you do it with a bunch of satellites, you have a bunch of spheres intersecting at one point, and that point is you position. In conventional earthquake location, the ranging measurement comes from an S minus P phase time delay run through a travel time curve, which gives you a range to a seismometer whose position you know. You do that with a bunch of seismometers and get intersecting circles, and that gives you a position. So it's a very similar thing, but with GPS positioning there are atmospheric dynamics, ionospheric dynamics, orbit issues, satellite clock issues, all that sorts of other errors that degrade your resolution.

There's also the issue of getting datasets freed up around the world, which is something I've been working on the last couple years. The pandemic really slowed that down, so we're about where we were in mid-2019 because these networks are very expensive to build and operate, and people often don't want to or can't, by national law, share their data. So ff you're trying to get Peru to open up its data streams, there are issues there. So I've been dealing a lot with that, while training graduate students and teaching classes. We have a master's program at CWU, so they only resident for a couple years. Typically, the best ones go on to be PhD students, where they finally use what we've taught them. Then, there's teaching. And that's kind of the job.

ZIERLER: Let's now go back and establish some personal history. As an undergraduate, was seismology already on your radar?

MELBOURNE: Yes, definitely. I grew up in Southern California, and earthquakes were things I had felt. I thought, "That's pretty cool." I went off to college as a physics major, and that was something I was very interested in. In college, I discovered the whole outdoorsy world of backpacking and climbing and mountaineering etc. I was exposed to a former Caltech faculty member, a guy named Brad Hager, and I realized he was doing math-y, physics-y stuff out in the world and got paid to go out camping and placing instruments. I thought, "That is for me. That is what I want to do." It was some combination of it being math- and physics-oriented as well as discovering backpacking and outdoorsy stuff.

ZIERLER: Where did you do your undergrad?

MELBOURNE: I started at Reed College in Portland, and Brad was the one who told me about a program where you could be a physics major up there, then transfer to Caltech, so I did that. And it was dynamite. Reed was really, really good at all those humanities. I still think the most important skill I have is the ability to write a proposal. I don't know how many literature classes I had, writing papers and more papers. Then, getting the basics as a physics major. I don't think that program exists anymore. Then, getting thrown right into the full Caltech undergrad thing, electrodynamics taught by Jackson, Feynman's math book, it was pretty hard. That's why I took a couple years off after college. "I don't want any more of that." [Laugh] But after a couple years of working, I was like, "Actually, I think I do want more of that."

ZIERLER: The program was two years at Reed and two at Caltech?

MELBOURNE: Three at Reed, two at Caltech.

ZIERLER: Is there a built-in master's with it?

MELBOURNE: No, you just get a bachelor's, but you get a good bachelor's. [ And I really came to value that. I remember when I first showed up at Caltech years later as a grad student, there was a professor of atmospheric science, Andrew Ingersoll, and he was in charge of giving out this test of math and science skills to incoming students. I took this test, and I just crushed it. I remember thinking, "Wow, either this place is easier than I expected, or I really got my money's worth out of my college experience." And that was where I think a lot of the Reed stuff helped. The chemistry classes there were hard. I'm not a chemist, but I really did learn that stuff inside and out. When they were like, "What's pH? What's electronegativity? How does this reaction work? What's the equilibrium constant?" I knew the answers. Even today, I'm not a chemist, but I remember that stuff.

ZIERLER: Did you get to hang out at the Seismo Lab as an undergrad?

MELBOURNE: Not really. There was no mixing. I'd see these people in the copy room, and I didn't really know much about it.

ZIERLER: And you knew you specifically wanted to take a break, not to go straight to graduate school?

MELBOURNE: Yes, absolutely.

ZIERLER: What were your options when you graduated?

MELBOURNE: I did like everybody else, I applied for a bunch of jobs. At the end, it was a stark choice. Very telling. One was a geophysics entry-level position at Unocal at their thermal plant in the Eastern California Salton Trough. They had a geothermal energy plant there. Unocal is oil, and the salary was just over the top. I had no idea you could be paid that much as a 23-year-old, so that was very compelling. The other job–you've got to put this into 1991 dollars, but I recall Unocal offering about $70,000 a year, and the other job was this USGS job as a GS5 up in Vancouver, Washington in the Cascades Volcano Observatory for $23,000 a year. It was decision time. [Laugh] But I noticed the Unocal people were very formal, and the USGS people were very casual. I noticed the Unocal people were in the Inland Empire way down in San Bernardino, and the USGS people were in Vancouver, Washington and got to fly on helicopters into volcanoes. You're 23, and you're not really thinking about retirement or any of that kind of stuff, so I took the USGS job. Maybe I'd be richer than Midas if I'd gone to Unocal, but who cares? [Laugh]

ZIERLER: Did you have an appreciation of the historic connections between Caltech and the Survey? Did that influence your decision at all?

MELBOURNE: No. I was just looking for a job. I knew I needed a job. I didn't have any money or income. I needed a job. [Laugh] I was looking all over. In time, I would come to appreciate those connections. The Seismo Lab has an extraordinary legacy. When you zoom way out and look at earthquake science, it's singular. I would say it's a place that's had a singular contribution to the discipline. But that's Caltech. I know it's like that with physics, AI, Carver Mead, Hopfield, all these others. They were doing the basic stuff on which all of our Google searches are based. But they were doing it 35, 40 years ago. That's one of the trippy things about Caltech. You get used to there being a Nobel laureate down the hall and going about your day, trying not to let it mess you up. [Laugh]

ZIERLER: Tell me about the work in Washington. What were you doing for the Survey?

MELBOURNE: It was crustal deformation. The Cascades Volcano Observatory was founded very quickly in response to the unrest at St. Helens. It was two months between when St. Helens turned on and when May 18, 1980 happened, as I recall. This was a ragtag team that was assembled really quickly by the USGS to monitor that volcano. After the big eruption, the volcano continued to erupt sporadically throughout 1986. Then, as they do, it kind of mellowed out for a while. By the time I got there in 1991, it wasn't really doing much. But we were measuring the deformation of the volcano. Even though it wasn't erupting, it was still very active. The measurements then were two types. One was a leveling, so you'd go up and down a road with leveling rods, which sounds archaic, but even today, in 2022, if you want only vertical deformation profile, the most precise way of doing it is still a leveling line. Good old-fashioned, 100-year-old technology. The reason GPS can't do it better is because of the incredible correlation between travel time through a variable-velocity atmosphere, and clock and satellite just destroy your vertical repeatability.

That was one measurement, leveling lines. As a GS5, my job was to hold the rod, then let the pro GS7s run the gun. [Laugh] The other one was laser pinging, two-way travel laser time. But you got to get in a helicopter, fly up into the caldera, and hammer one of these mirror reflectors onto some steaming-hot wall of rhyolite, and then get out of there. I enjoyed that immensely. Nobody was allowed into the caldera of St. Helens unless you were USGS. Once I looked inside one of these cracks in the rock on which I was attaching a mirror reflector for the laser system, and it was glowing red-hot in there, just a meter below the surface. We were drilling a hole, putting a reflector on there, hoping nothing happened. You didn't spend a lot of time inside the caldera in 1991. The chopper would drop you off, you'd do your thing, and you'd get out of there really quick. Maybe I'm atypical, but that was a very exciting job. [Laugh] I was at it for just over a year, maybe 15 months, but in time I realized the avalanche of learning that had burned me out pretty totally in college over the past five years had stopped, and this lack of learning is what caused me to want to go to grad school.

I don't want this to sound in any way like an indictment of the USGS, but I was like, "I'm not done with this learning stuff process, so I'm going to sign up for more punishment." I kind of realized I wanted to keep getting more knowledgeable. And maybe this is reflective of me, but in '92, the web hadn't been invented. You couldn't just get interested in something and go find out about it the way you can today. So I realized I wasn't done learning yet, and I wanted to go back to school. I only applied a few places. Joann Stock felt strongly that , "You've got to go back east." I was, like, "I'm not going back east." So I applied to Caltech, Scripps, and the University of Washington. And I didn't think Caltech would take me honestly. Seemed like a bad idea, I don't know, to go back to the same institution. But they did. And as I planned to go into seismology, given names like Benioff, Gutenberg, Richter, Helmberger, Kanamori, on and on. I thought, "I would be an idiot not to go there." So that's how I ended up going back after my time at the USGS.

ZIERLER: Did you have a good idea, since you were knowledgeable of the department already, who your thesis advisor would be?

MELBOURNE: No. Caltech's really special that way. I don't know how the faculty sort it out, but they just sort of do. I wasn't beholden to anybody. Joann was so gracious to let me work on her Jalisco project as an orals project, then I was like, "Hey, I'm going to go do this other thing," then when she had a big earthquake, she let me back to work on it. I've not heard of that happening any other place except Scripps. I think Scripps does it that way, too. I don't think it's a function of money, I think it's a function of faculty values, which really makes it an amazing thing. Again, I hope it's still like that- I've been gone for over two decades- but I think it is.

ZIERLER: Tell me about the Jalisco project and what was compelling to you about it.

MELBOURNE: Well, it's fun because as a first- and second-year graduate student, you're learning a lot of things simultaneously. I was a C programmer of sorts, but not in a Unix environment, so I was dealing with Unix, learning how not to cat binaries to a screen, all that. You're going through this massive learning curve, eventually get the hang of it, and things start to make sense. You're parsing numbers, writing scripts to do little things. There was a fairly new at that time plotting and map-making package called Generic Mapping Tools, and I was learning how to make plots and maps, and learning the GPS processing software, and a bunch of other things simultaneously. At some point in all this, after days of work, I finally made a vector map of how things moved before and after the earthquake on all of our Jalisco GPS stations. I remember the first time it popped up on the screen, it was so beautiful. It was like mother nature just tugged on the crust of Jalisco and pulled it out to sea, all 13 stations just moved as if the crust was made of rubber. The way this popped out of weeks and weeks of crunching numbers, for me, was probably the singularly coolest moment in my graduate story. I remember parading it all around the third floor of the Seismo Lab saying things like, "Look, it worked." And that was from 10 months before the earthquake and six days after.

But when you looked at days six and seven, seven and eight, eight and nine, nine and ten, and we went out to eighteen days or something like that, the crust just kept moving many centimeters, well within our resolution, which meant the fault was still sliding. What does that mean, when you have an earthquake, but it doesn't turn off, it festers for days and days? That was the beginning of, "Earthquakes aren't just this thing that happens, they're part of a very dynamic landscape or spectrum of behavior." That was cool. We could map out not just where the–you get these offsets, and if you think you know where the fault is, it's just a matrix inversion to figure out what distribution of slip best predicts those offsets you measure. Then, you can do that again over time, you can see these slip patches propagate around.

As they propagate sideways, they're kicking off aftershocks that pick up on the seismic network, that was far and away the coolest aspect of the project. Driving around back roads of Jalisco was also very interesting. Jalisco is a fascinating part of Mexico very much unlike other parts of Mexico. Mexico once had a thriving film industry, so if you've ever seen old 1940s Mexican films, chances are they were filmed in Jalisco. Jalisco also owns the trade name to tequila, so that was cool, too, a lot of distilleries out there to visit while driving around. I think today with the narco traffic problems these days, you probably can't or shouldn't get into a lot of these places we went to. But boy was it neat then. Of course, all this tequila distillery business took a distant second to the science itself.

ZIERLER: More broadly, between coffee hour and interacting with fellow students and faculty, what were the big ideas? What was animating people at the Seismo Lab when you were a grad student?

MELBOURNE: Because Hiroo was far and away the most religious attendee of coffee–it was his coffee break, for coffee, so it was oftentimes what he was interested in. During my time, he was very interested in fault fluids moderating or triggering earthquakes. If fault fluids existed in a fault at the overburden pressure of a rock, then they're prying the fault open, which should enable it to slip. There are lots of reasons to think this is the case, even to this day. Hiroo was very interested in that. Helmberger, who ultimately became my thesis advisor, had discovered with his other students these ultra-low-velocity zones on the core-mantle boundary, which were very detailed but incredibly low, like, 20% below reference velocity. ULVZs were a big deal, fault fluids was a big deal. Joann had a lot of students going research cruises off of Antarctica at that point, mapping out the plate-tectonic history of Antarctica vis-a-vis the Southern Indian Ocean or something like that. And she needed those because she was working on these big global-plate reconstructions, and a lot of uncertainties in the global reconstructions pinned down to what was going on up there. There were always students coming and going, telling stories about getting trapped in the Drake Passage between the Antarctic Peninsula and Southern South America in 110-knot wind or whatever. Rob Clayton was running all kinds of seismic lines. LARSE, I think it was called. Los Angeles Region Seismic Experiment or something like that. There were all kinds of seismic reflection lines being run all over LA. I got to help out with one, which was kind of fun. I recall being in the ‘hood, my station, at 2 in the morning. I would say those were the big themes.

But I do have a mild criticism of that time. The Seismo Lab faculty had not kept up with computer times, so there were a couple years where we didn't have the computer power we needed. They were trying to keep an old mainframe going, a Solbourne, I think. Meanwhile, Linux had come around, and we students just blew right past the faculty on this. We were all going out and buying and building our own PCs. We'd go to flea markets to get motherboards, disc controllers, hard drives, put them all together, and put Linux on it. The resulting computers were far faster than what the Seismo Lab had available to compute with. But computers are a rat hole that you can go down that get in the way of doing science, for sure, and boy, I was surely at risk. That would've been about 1995, '96. Today, computers are the least-expensive piece of the research equation, and of course, everybody's got their own MacBook Pro now that's a supercomputer. So that was a big issue for us, we were building our own computers and recompiling endless software packages, all at great time cost, because the computer system in the Seismo Lab sucked. [Laugh] But I mean that as only a mild criticism.

ZIERLER: You mentioned computers. This is going to sound like a long time ago, but what about the internet? Was the internet fully embraced for seismology at that point?

MELBOURNE: Yes, definitely. A lot of the seismic data, we just downloaded it from the data management center, which was run by the NSF through a contract. But the World Wide Web came in right in the middle of my graduate time. Mosaic was the first graphical browser, and that changed everything really quickly. I think a lot of the Seismo Lab PhD graduates who didn't end up sticking with pure science went off to Google or the big software houses. I know a few from my year.

One other thing I would point out is, and this is something I've struggled with a lot, I've asked myself many times, "Was Caltech good for me?" The overwhelming empirical evidence says, "Yes, it was incredibly good for you." But there's so much stress of a graduate student that you're just not good enough, not smart enough, not working hard enough, not productive enough. And in time, I came to understand that the dimensionality of human cognition is vast, and that odds are, you're going to know something that somebody else around you doesn't know, or have a good idea that nobody else will ever have. But at the time, it seems that everybody there except the faculty felt inadequate. At some point I compared notes with other students who have said, "Oh, Yes, we all felt that way." I think today that maybe that is just the baptism by fire that you have to go through to mature scientifically. And we all kind of hid it the best we could, but I've often thought, "I wonder what that did to my psyche." I don't mean that in any disparaging way whatsoever. I have such extraordinary sense of thanks for the entire Lab, all the people who were there, because you actually learn as much from your fellow students as you do from advisors. But I won't say it came at no cost. It was, and I suspect still is, very stressful. It's a bunch of motivated people who are working hard.

ZIERLER: How much of that was inward-directed, you competing against yourself, and how much of it was as you looking around at the other graduate students?

MELBOURNE: 90/10. Most grad students were helpful, though there were always a few exceptions, but I certainly always tried to help anybody if I could. There were people who knew more about computers, for sure, most of whom were happy to share their knowledge. Mostly, it was just a pervasive sense of inadequacy [Laugh]. I had a lovely escape, though. When the stress got too bad, I'd just go for a run. I'm a runner, slow and middling, but dedicated. And I was like, "You got the Arroyo right there. Get in your car, run 10 miles in the Arroyo, come back, and things won't seem as bad." Always worked like a charm, still does today. But Caltech was a very stressful place, at least for me, particularly at the beginning of my graduate school.

ZIERLER: For the research, when did you know you had enough to defend? When did you feel like it was a complete project?

MELBOURNE: This is really interesting. It just hit me one day. I think I'd been there five years at that point and I was like, "I'm done. I know what I'm doing. I need to get out of here." At that point, I was working entirely with Don Helmberger, we'd published three papers, not counting any of the Jalisco earthquake stuff I'd done with Joan. So I was said, "Don, I'm going to get out of here. I've done enough." He said "Yes, fine, I agree." I remember thinking, "Why didn't you say something earlier?" Of course, from his standpoint, I'm now the most productive now and he had no incentive to see me move on. I now feel the same way about my senior grad students who are finally getting stuff done. My papers with Joann on the Jalisco earthquake were already published, so there was no discussion there. So one day, it just kind of hit me, it was time to move on. I don't want to say I didn't need those people anymore because that sounds so mercenary, and that wasn't the sentiment. It was more like the eight-month-old eaglet in the nest beating its wings just had to take that leap. I felt "I'm in control. I know what I'm doing, I know why I'm doing it, I know how it was done, I wrote every one of these papers. I wrote or know the people who wrote the codes I'm using." The phrase I used at the time was, "I feel like I've got critical mass between my ears to go do meaningful stuff." And I didn't feel that way at all until I did. It was about five and a half years for me.

ZIERLER: What did you see as the main contributions or findings of your thesis, and how have they aged over the years?

MELBOURNE: Nobody's disproven any of them, so I'd say they've aged well. Might be a bad sign that nobody bothered because it's not worth it. [Laugh] Clearly, the advent of continuous geodetic measurements enabled the realization–I remember my thesis-defense seminar, talking about this feedback between fast and slow slip. One triggers the other, the other triggers the other, and this is how you get chaos. Chaos theory was still popular at that time. What Joann, I, and our colleagues teased out of the Jalisco dataset has turned out to be absolutely true. The San Andreas Fault, for example, is not just this crack in the ground that's not sliding. There are all kinds of activity going on in the deeper reaches of it. You've got to be able to measure it and get in there with fine measurements to see it, but it's active. But that realization that these faults are almost alive with activity and doing stuff continuously, and that the big one is just one piece of the total spectrum of how they behave through the decades, that came out of my thesis for me and the papers we published, and this sentiment is now widely accepted. There are a few big faults in the world that seem to be truly and totally locked, but they're the minority.

ZIERLER: Who was on your committee besides Helmberger?

MELBOURNE: Tom Ahrens, Don Helmberger, Rob Clayton, Joann Stock, and I think Hiroo was on it. But I may be confusing it with my orals committee, which was scary because that was the one where they may kick you out. You're not going to defend your PhD if your advisor doesn't think you're done. So if your advisor thinks you're done, and I think I ended up having six papers published to peer review, that's going to be one hell of a committee that's going to hold you back. [Laugh] I hope this doesn't sound wrong, but becase of this, the PhD defense committee is just a rubber stamp, as far as I'm concerned, whereas the orals exam in year 2 is the one with teeth- they can kick you out. That's why I'm struggling to remember who was on my Phd committee.

ZIERLER: Anything memorable from the defense, questions or debates?

MELBOURNE: Somebody asked why I hadn't modeled the dynamic feedback between slow and fast slip triggering each other, and I forget who, but I was like, "That's another PhD thesis unto itself with a very different set of skills, disciplines, and theories." I remember thinking, "Should I have done that?" Today there are whole divisions working on that problem.

ZIERLER: After you defended, were you looking at faculty and post-doc opportunities at the same time?

MELBOURNE: I was. I had a post-doc at MIT, and I had what I was told was an incipient faculty offer at University of Washington but they were being slow, a faculty offer in hand at Central Washington University, and I think there was the making of a staff scientist position at JPL. But I had a kid on the way at this point, so the idea of a post-doc wasn't as appealing as a permanent job. The incoming baby had a huge impact on my decision-making. I was also very interested in using geodesy to monitor earthquakes and subduction zones, and I knew this network was getting built at CWU, and that was more interesting to me scientifically than what the MIT post-doc offered. So I was thought, "All right, I'm going to go where this network is starting up at this non- Research-1 university and see how it works."

ZIERLER: As you've already told me, quite well.

MELBOURNE: It worked out. As somebody at CWU told me, "You're not going to maximize your scientific productivity at CWU. If that's what you're gunning for in life, don't come here." I thought hard about it, and today, I think that is true. If you don't have PhD students, it's very hard to get papers across the finish line, and you are limited in what you can take on. I think if I had gone to Berkeley or wherever, I probably would've had more going on professionally, certainly published more papers and whatnot. But I don't think it matters, at least for me. It kind of depends on what provides you affirmation in your world, what's meaningful to you in your world. There are times in my job where I definitely get frustrated that I can't do more stuff, take on more things. I'm pretty good at seeing around scientific corners, it turns out, and there are some things that have come along where I knew, years out, what fruit would be there to have. If I'd had post-docs and the whole package of a Research-1, I would've been all over it. Instead I've had to be very judicious and strategic about what I choose to take on with just Master's students and a few staff. On the other hand, I live in a beautiful place with a river running though town and Bald Eagles fishing out my window, I walk to work, and there's nothing but national parks and forests on all sides. I live on the east slope of the North Cascades, an hour and a half from Seattle, and that's got measurable benefits as well. So when you add it all up, there's no perfect solution on the formula for a good life. But if I could move the Seismo Lab to CWU, I most certainly would!

ZIERLER: I think I'd follow you. I'm missing the rain in a very big way. [Laugh] What about the funding? In other words, are you at a structural disadvantage not having PhD students when you're applying for grants?

MELBOURNE: That's a tough question to answer. I don't see how I could not be. The last four years of a PhD's typical six-year residence are the productive ones, and if you can use those figures to write your proposal, it's so much easier. There were times where I was like, "I just can't do this. I can't make this work here. I can't go up against the Caltechs and Stanfords of the world." But I think there's enough to do, enough creativity in the process of science, that even if you don't have all of that infrastructure, if you can write, and I've got that box checked, you can still make the case for meaningful scientific things to be done that are not going to come out of other groups. They have great ideas you'll never have, but you'll have great ideas they'll never have. If you can write and pitch those great ideas, the funding agencies seem to go for it. I've often thought, "The funding agencies largely report to Congress, and Congress is going to get pretty irritated if they only fund UC, Caltech, Stanford, MIT, and Harvard. They've got to spread the wealth a little bit." I think maybe what I lack in the infrastructure of a PhD school, maybe there's a bit of a tailwind from the sense of trying to spread the wealth of American science funding to non-Research-1 universities. It's a hard question to answer, but I keep in touch with Caltech and other universities' faculty, and sometimes I see great stuff going on there, and I'm feel, "I'd love to be able to do some of that stuff." But if you can't, you can't. It's a little bit like running a marathon. If you can do a five-minute mile, fine, go out at five min miles and hope you can hold it for 26. If you're a ten-minute-per-mile shuffler, know that about yourself, plan accordingly, and you'll probably run a great marathon. These are not good answers, I realize, but I guess I don't know the answer to your question. I feel structurally, there's no way you can really compete, yet somehow, going on 24 years, I have, at least by some metrics.

ZIERLER: Maybe the answer is, in fact, yes.

MELBOURNE: Yes, maybe. [Laugh]

ZIERLER: To flip that around, specifically as a product of Reed, being at a small college where teaching is really emphasized, what's the value in that, given that you have that bandwidth to interact with undergraduates in a way that might not be possible at a larger research institution?

MELBOURNE: Let me give you an example. There was a humanities class we had to take at Reed. You read The Iliad or whatever, write an essay, turn it in. You get it back, graded. Then the professor sits down with you and goes through that essay with you, paragraph by paragraph. Then, the next week, you rewrite it. Then you rewrite it again. And again. You do this for 13 weeks straight. Eventually, seven or eight weeks into this, you've got this perfected essay, but it's kind of boring. That's when the professor sits down with you and says, "The nuts and bolts are there, but it's really quite boring. Why don't you bring in what's going on today in national politics?" I remember at Reed, there was a lot about old-growth timber, the spotted owl, and stuff. "Bring that into this essay. Use this essay to say something about the world we're in today" You do this, and suddenly, by the end of this course, you go back and compare the 13th version and the first version of it, and what you've got at the end is something you might submit to the New Yorker. It's really interesting, and that's how one really learns to write.

That one-on-one, something like that could happen at Reed because that's what the place is all about. This is the gold standard of education, and you come out of it knowing what the best education is and how incredibly expensive it is to deliver, because of it's one-on-one nature. I take that lesson with me when I'm teaching our undergrads at CWU. Then, the Caltech training of world-class science, the process, the integrity, the values, and the professionalism that goes with world-class science, that comes through, too. You really get science excellence drilled into you at Caltech. I think in my dealings with my students, those values and experiences, the gold standard in both education and scientific training, inform everything I talk to students about.

I can give you an example. I had a very, very talented grad student from Costa Rica, and we were studying these slow-slip events. I was in charge of the GPS stuff, but she wanted to do the seismology. I pulled her out of OVSICORI, which runs Costa Rica's national seismic network,, when I taught a class down there. She came to CWU to do seismology, not GPS, and was looking for a thesis. I told her about the discovery in Japan of the seismic signature of slow earthquakes, colloquially known as tremor. "This guy, Obara, has discovered tremor. Let's start looking at tremor in these events in Cascadia."

I remember telling the student, "We're going to need to have a wall between us scientifically because we're going to compare what you're seeing on the seismic networks with what I'm doing geodetically with the GPS, and our comparison has to be a double-blind. You cannot be seeing what I'm doing, and I cannot be seeing what you're doing." And she and I would establish the extraordinary linearity of moment release in these tremor events with respect to their duration, and nothing else. But to do this right, it was very important to keep the double-blinds up.

This emphasis, which I passed on to her, came out of coffee discussions at Caltech about the important data is the data that didn't fit your model. You do not throw misfitting data away, that's the most important data you've got because it tells you your model's not doing something, not explaining something. I think I'm mangling my answer here, but somehow between the pure–Reed education standard and the Caltech science standard, when I deal with own students, both those two institutional voices and values are in my head, guiding how I think and what I do. I should add, I'm very, very lucky- as I get older, I realize how incredibly fortunate I was in terms of the training I received.

ZIERLER: For the last part of our talk, a few retrospective questions, then we'll end looking to the future. What has stayed with you from your training at Caltech? What informs your approach to the data, the way you collaborate, your sense of the important things to focus on?

MELBOURNE: There's a whole bunch. I would start with something I heard both Don Helmberger and Hugh Taylor say. "Stick to the data." Models come and go, interpretations come and go, but the one thing that doesn't change is the data. The closer you stay to the data, the safer you'll be. That's always been there in my head, stay close to data. Another one is harder to put words to. Life throws curveballs at you of all kinds. I've had a few. But the idea of me, as a scientist who should be spending most of the time thinking like a scientist about stuff, it's an attractor that brings me back to who I am, and I think that's a function of Caltech. I can get interested in something for a while, but eventually I come back to science. I need to be looking at data, thinking about the world as a scientist, expanding this, trying to do that, model this, or whatever.

Probably if I hadn't gone to Caltech and gone through all that, I'd be chasing something else, money probably. I've come to think that making money is probably what fills that void, mission, purpose, in cognition for a lot of people who don't have some other passion, like research. I assume that's how it would've been, I don't know. But for me, it's science. Do stuff that's scientifically meaningful. That may be limiting. It doesn't mean I'm doing something meaningful in terms of stopping world hunger or global peace. But stay true to science, stay a scientist, and educate. Increasingly, as I grow older, making sure these students kind of get it is incredibly important. That's a growing principle for me, and I attribute that entirely to Caltech.

ZIERLER: On the note of the science specifically, what were the open questions when you were a graduate student, and among them, what was resolved, and what perhaps remains as open as when you first encountered it 25 years ago?

MELBOURNE: Whether earthquakes are time-predictable or recurrence interval-predictable was still being bandied around. I think the preponderance of evidence says no, they're not. Nobody debates that anymore. One thing that hasn't changed, which is infuriating to me, is some of the fine detail of these big structural transitions deep in the earth. Transition-zone structures, core-mantle boundary structures. This is probably going to irritate all kinds of Seismo Lab faculty. The earth attenuates the high frequencies, absorbs them. But it's the high frequencies that carry the precise information that one would need to push the discipline forward, but even if you could model them, the earth absorbs them so there's nothing to model! In volume three of the Feynman lectures, he's got a very beautiful explanation of how, when the diffracted energy of wave propagation starts to rival the direct energy, you lose all resolving power. Feynman uses it in talking about why, with optical microscopes, you can only see so small, then you've got to go to electron microscopes with much shorter effective wavelengths.

In seismology, if you're losing the high frequencies, you're losing the detailed statements about the physics ability to resolve detailed structures. So long as earth continues to absorb high frequencies and preferentially transmits the lower frequencies, you're never going to have that information. There are a lot of details about deep structures of the earth–it's infuriating because a year or so ago, I was asked to chair an AGU session on the exact same stuff that Helmberger and I were publishing in the mid-90s. That discipline hasn't changed in 30 years! I was like, "Screw this, man." [Laugh] There are some things like that that are just limited.

But for me, the big topic for of my career is the incredible activity ongoing along active earthquake faults that aren't currently undergoing big earthquakes, but are nonetheless doing all kinds of other interesting things that only the newest technologies has allowed us to see. I'd say those are the big ones.

ZIERLER: Finally, last question, looking to the future. However long you define a research plan or agenda, 5 years, 10 years, whatever it is, what's left for you to accomplish, and how do you get there?

MELBOURNE: I've actually got a piece of paper on my desk, and whenever I have an inspiring idea for a research project, I write it down. There's a lot of stuff on that paper these days. I would like to pick off three or four of these items before I retire. I might have 15 years left in my career, maybe 20. A couple of them are fairly meaty topics. For example, I don't believe anybody in the greater discipline of seismology has really combined extremely broadband seismometers, ones that can easily see solid earth tides, storm fronts approaching, things like that, with the ITRF, which is the geodetic reference frame we use in GPS, so that they're in a common reference frame. Do that, and then start looking at some of these slow-slip events we see in Cascadia, and see if we can pull out 500- or 1,000-second-long slip pulses that elude strainmeters, are too small for the GPS to see and are so far out of the spectrum that the Streckeisens can see, typically, unless you orient them in a reference frame because you can't correlate the signals from station to station. That's a project I've been chipping away at for the last year or so.

Another topic is I would like to see this GPS-based global seismic network we've built that comprises thousands of high-rate GNSS receivers from around the world capture a big earthquake and characterize it in 20 or 30 seconds and use that to inform either a tsunami or strong ground motion early warning message. That would be quite an affirmation. We did capture the California Ridgecrest earthquake in 2019, but that was small, a 7.1, I think. Capturing one of the big M8 events around the world would be really gratifying for me, since we've put so much time into developing the system.

ZIERLER: That's technologically feasible, just a matter of being in the right place at the right time?

MELBOURNE: Yep. We're doing it right now, and all the filters are running. We just need the earthquake to happen under a subduction zone where we have access to GPS receivers. We have 1,800 coming in around the world. I'm rooting on Chile, Alaska, or New Zealand. It would be affirming because if I'm two-thirds of the way through my career, you start to look back like, "Wow, okay. I hope I made something of it." Seeing the system put to use to save lives would be very gratifying. The system does work well, so it's just a matter of waiting.

ZIERLER: On that note, this has been a terrific conversation. I'm so glad we connected for this. I'd like to thank you so much.

MELBOURNE: Okay, thank you.