This transcript was generated automatically and lightly edited for readability.

Oliver Strimpel: This is Geology Bites with Oliver Strimpel. It's not very far away from us, perhaps less far than we would drive to go shopping, but the direction is down, straight down. Yet the farthest anyone has ever reached is 12 kilometers below the surface. This is a tiny pinprick on a planet whose center is almost 6,400 kilometers below. So how can we figure out what's going on deep inside the Earth? Barbara Romanowicz is a professor of the graduate school at the University of California at Berkeley, and chair of physics of the Earth's interior at the Collège de France in Paris. She uses the seismic waves triggered by earthquakes to probe the interior of the Earth. Using her training as a mathematician, she has forged new techniques for analyzing these waves to give us a much sharper view of the deep structure of the Earth. Barbara Romanowicz, welcome to Geology Bites.

Barbara Romanowicz: I'm looking forward to it.

Oliver Strimpel: As I just mentioned in your introduction, you've sharpened our vision into the Earth. But before we get to that, how do we see into the Earth at all, to begin with?

Barbara Romanowicz: Well, we use seismic waves to illuminate the interior of the Earth, just as light waves illuminate what we see around us at the surface. Seismic waves travel at speeds that depend on the material that they traverse, and so by measuring their travel times, we can get information about which materials they've traversed and what their physical state is: how hot they are, what their composition is, and other properties.

Oliver Strimpel: So if you measure the travel time, what does that tell us?

Barbara Romanowicz: Well, the travel time is directly related to the distance that the waves have traveled between the source, which is a natural earthquake, and the seismic station, which records the wave field. We can figure out what the path of the wave is, because we have reference models of the Earth that allow us to do this, using some simple principles like Fermat's principle and the laws of refraction and reflection, the same way as we do for optical waves. And so just dividing the distance by the travel time will give us the average velocity along the path.

Oliver Strimpel: So do we assume that the waves travel in a straight line between the earthquake and our stations?

Barbara Romanowicz: No, they don't travel along straight lines. They travel along curved lines, due to the fact that the velocity increases, on average, with depth in the Earth, because of pressure inside the Earth, which makes the material more and more compact, and therefore more capable of rapidly transmitting waves as they go along. So as they dive into the Earth, they will gradually get curved back towards the surface, and that is how we are then able to see them at many, many different points around the Earth.

Oliver Strimpel: Just give me a feel for it. Roughly how long does it take, say, for a wave, if you have, say, an earthquake in Japan and a sensing station that you might have in Berkeley, where you're based?

Barbara Romanowicz: Well, it will take a couple hundred seconds. Seismic waves travel at several kilometers per second inside the Earth. You know, depending on the distance, you will have to wait longer and longer, but it is several hundreds of seconds to several thousands of seconds, depending on the waves and the distance.

Oliver Strimpel: Hmm, so we're talking about minutes, really.

Barbara Romanowicz: Yes, we're talking about minutes for large distances. Of course, if you're right near the earthquake, it will be seconds rather than hundreds of seconds. So it all spreads out with the distance.

Oliver Strimpel: In addition to the bending of seismic waves as they go deeper, do they also bounce off, or reflect off, sharp boundaries in the Earth?

Barbara Romanowicz: Yes, you have both situations; it depends on where you are in the Earth. The surfaces that are the most reflective are the surface of the Earth and also the core–mantle boundary, 2,900 kilometers down from the surface, where we have the juxtaposition of a solid, the solid mantle, and the fluid core, and that's a very good reflector. So we see a lot of waves that reflect from the Earth's surface and from the core–mantle boundary. And then there are some secondary discontinuities inside the Earth's mantle, for example one at about 660 kilometers, which corresponds to a phase change, a change in the crystal structure of the rocks, of the materials. And that one is also capable of reflecting waves, but the reflection coefficient is much smaller, so the reflected waves have very small amplitudes. You have to devise special methodologies to pick them out.

Oliver Strimpel: That makes me wonder: how do we know which path the wave took to get to us?

Barbara Romanowicz: Well, this knowledge has accumulated over the years. In the first half of the 20th century, seismologists progressively were able to build very simple models of the Earth's interior, where the Earth is made of spherical shells. And progressively, by including more and more observations, they were able to get reference models good enough that we can, by our calculations, predict the arrival time of a particular seismic wave and then confront it with the observations. And generally the differences in arrival times are a fraction of a second to several seconds, and often not ambiguous. And also, if they are ambiguous, we can distinguish them by looking at the polarization of the waves, because there are two types of seismic waves, compressional waves, called P waves, and shear waves, and they have different polarizations. So you can also use that information, if you can measure their arrival on three components of motion, to basically identify them.

Oliver Strimpel: But if you can measure the travel time, even if you know which path a wave took, that still only gives you the average velocity for the whole distance. So how can we tell what was changing along the way? Like, it could have gone very fast for the first bit and slow for the second bit, or slow for the first bit and fast for the second bit, and it would look the same in terms of total travel time.

Barbara Romanowicz: Absolutely. You cannot tell from just one recording, from one wave going from one source to one receiver, one type of wave. You need many recordings, at many places around the world and from many earthquake sources, and it's only by combining all of these and back-projecting your observations, which are, in general, in terms of travel time anomalies compared to what you expect from the reference model. And so the travel time anomaly will be positive if the wave has been slowed down somewhere on its path, and it will be negative if it's been sped up. And then, by comparing all of these recordings, you can locate where these anomalies occurred and also determine their strength. But you need all of these many, many different paths to illuminate the interior and get that kind of information. You could do it by hand if you wanted, but it would take a very, very long time. So you turn it into a set of linear equations, and you solve that set of equations to give you the answer.

Oliver Strimpel: So you really need to have a model of the Earth to disentangle what happened to a seismic wave as it traveled through the Earth, and then the observations will either confirm the model or, more likely, make you tweak it a bit, I suppose. I suppose you'd call this an inversion problem.

Barbara Romanowicz: Yes, it's an inverse problem. You measure, say, the travel time, in the simplest case. You measure the travel time of a particular type of wave, for example P waves, the first-arriving P wave, at many different stations for many different earthquakes, and you try from this set of travel times to infer the velocity distribution in three dimensions inside the Earth. It's called seismic tomography.

Oliver Strimpel: So that makes me think at once of X-ray tomography in the medical context. So is this the same basic inversion technique that's used in the medical context?

Barbara Romanowicz: It's very similar. The concept is the same. What is not the same is the observation that is used. In the case of seismic tomography, it is primarily the travel time of different types of waves. In the case of medical tomography, it can be other kinds of observables; very often it's the attenuation of some kind of wave field, depending on the technique. But the idea of tomography, of observing at the surface of the body and inferring its internal properties by an inverse procedure, is the same. And so the techniques, which were, I believe, developed first in geosciences, have been ported to the medical field and improved tremendously in that field, because they have the advantage that they can put sensors and sources all around the body. That is not the case for us, because we depend on the natural earthquake distribution, which is not at all uniform around the Earth; it's mostly along the boundaries of the tectonic plates. And then we also depend on the distribution of stations, which are mostly on land, right? And the Earth is two-thirds covered by oceans, so that's a big problem. We have a lack of sampling, basically, that we cannot do anything about, except improve the sophistication of our methodologies.

Oliver Strimpel: Unlike the X-ray medical application, don't the geologists also have a different problem, because the speed that we're measuring depends not only on temperature, which is interesting, but also on the composition of the rock it's going through? So how do you disentangle those two things?

Barbara Romanowicz: Well, to first order, temperature is the dominant factor, so you can always start by interpreting your velocity field, your images of velocity inside the Earth, in terms of temperature variations. Composition does play a role, and if you really want to disentangle it properly, you need to combine measurements for P waves, for compressional waves, and measurements for shear waves, which have different velocities, and, if possible, direct information on density, which is very difficult to get from seismology; you have to get it from geodesy, from measuring the gravity field and that kind of thing. And then, once you have these combinations, as much as possible, you confront this with mineral physics experiments, and now also theoretical computations. So mineral physicists will study the properties of rocks at high pressure and high temperature, at the pressures and temperatures that we know exist in the Earth, and they will be able to tell you how to interpret different velocities in terms of composition and temperature.

Oliver Strimpel: So what do we wind up actually seeing down there, using these techniques that we've just been talking about?

Barbara Romanowicz: Well, we see patterns of variations of velocities that we can interpret, let's say, in terms of temperature, to first order, and that will inform us on the dynamics of the Earth's interior. So specifically, if we talk about the Earth's mantle: the Earth's mantle is solid, but it does flow over geological time. It flows at very slow speeds, say several centimeters per year, like plate speeds, and the waves, of course, see an instantaneous picture of this and can inform us on where the upwelling currents are and where the downwelling currents are, of the convective system inside the mantle that drives plate motions.

Oliver Strimpel: So you can actually see the mantle convection cells that drive the motions of the tectonic plates riding on top.

Barbara Romanowicz: Well, that's what we're trying to do. We still have a pretty blurry vision of this. The parts that are best imaged are the subducted slabs. The plates get formed at the mid-ocean ridges, and then they get back into the mantle at subduction zones. At that moment, they have cooled down; they're cold, so they are manifested by fast seismic velocities, and we can follow them down quite deep into the mantle, at least to the boundary between what we call the upper mantle and the lower mantle, which is at 660 kilometers depth. This refers to the discontinuity that I mentioned at the beginning. And then we can kind of follow them deeper; it's a little bit harder. By the time we get to the bottom of the mantle, what we see is a ring of fast velocities surrounding two large areas of slow velocities. And we think that this ring of fast velocities is basically the graveyard of slabs, where all the slabs end up.

Oliver Strimpel: I'm impressed that a slab that was an ocean plate, and then went down, subducted, broke off, can retain its identity. It doesn't melt away like a block of ice would in boiling water. It lasts all the way down to the bottom of the mantle.

Barbara Romanowicz: Well, not continuously. Not continuously. Down to 660 kilometers we can follow them continuously, in fact; they don't necessarily break off. You can still see slabs that are active today, plates that are subducting, like beneath Japan, for example, or beneath South America. You can follow them down all the way to 660 kilometers as more or less continuous features. Below that, they have very frequently been broken up. You only see sort of blobs, because the patterns have changed over time and some slabs have been broken up. Subduction zones, the trenches, have retreated, and so the morphology has changed, and this creates breaks in the slabs, so you can't follow them continuously. Besides, as you said, they will heat up as they go down, and as they heat up, it's harder and harder to distinguish them from the surrounding mantle, obviously.

Oliver Strimpel: So do we actually see the slabs in the places that we expect them to be? For example, we know that the Tethys Ocean was subducted and pulled India north before it collided with Asia. Do we see a bunch of Tethys slabs on their way down below that part of the world?

Barbara Romanowicz: Yes, absolutely, we do. We see that very clearly under India and Tibet, under the southern part of Asia. And in fact, some researchers try to use this information to infer the past history of subduction, to infer the past configuration of plates. Of course, you have to make some assumptions on whether the blob that you're seeing does correspond to a particular time in the past, so you have to make assumptions about the velocity of sinking of the slabs. But these reconstructions, going back probably at least 200 million years, have been quite successful, in Central Asia in particular, for the Tethys that you're referring to, but also in North America, with the so-called lost Farallon plate, which we can see all the way down, broken up, not continuously, but we can see it all the way down to the core–mantle boundary under North America.

Oliver Strimpel: So this was a plate that subducted under what is now California?

Barbara Romanowicz: Yes, roughly speaking, yes.

Oliver Strimpel: On a smaller scale than the subduction zones, there are mid-ocean volcanic islands, such as Hawaii. It's thought that they are the surface manifestations of giant plumes of hotter material, called hotspots, originating very deep in the mantle. Can we actually see those with seismic tomography?

Barbara Romanowicz: Well, hotspot volcanoes are not explained by plate tectonics. They are mostly in the middle of plates, like Hawaii, and so the question of where they come from has been a subject of controversy for over 30 years, with the proponents of these fixed hot plumes of material originating somewhere down in the mantle opposed to the proponents of this being a local effect of stresses in the lithosphere. I think by now tomography has really made enough progress to be able to confirm that these hot plumes of material do exist, all the way from the core–mantle boundary, under some of these hotspot volcanoes, in particular Hawaii, but also Iceland, and other hotspots in the middle of the Pacific Ocean and in the Indian Ocean.

Oliver Strimpel: Coming back to the techniques and improving them: at the moment, with the methods you've just been describing, what is the spatial resolution of the imagery?

Barbara Romanowicz: The overall global-scale resolution with standard tomographic methods is, in the lateral dimensions, about two to three thousand kilometers, maybe 2,000 kilometers at the most, which is still too large, and in depth, a couple hundred kilometers. With new techniques that have been developed in the last 10 years, essentially, this resolution can be improved and brought down to, at present, I would say five to six hundred kilometers laterally.

Oliver Strimpel: What do these new techniques involve?

Barbara Romanowicz: The new techniques try to utilize not only the first-arriving waves, not only one single type of wave, like the P wave or the S wave. They also try to utilize the complete information in the recording of an earthquake, which includes many waves that have bounced off the different discontinuities and the objects they have encountered in the mantle. So by including the whole information in the seismogram, we hope to be able to compensate for the fact that we don't have earthquake sources and recording stations uniformly distributed around the world. If we did have that, then we could gain very high resolution with very simple methods.

Oliver Strimpel: So if you just take the first arrival, you get about 2,000-kilometer resolution. If you can use the whole wave train, which gives you all the different waves, the paths of the rays that got there, that gives you more information, and you can maybe do about three to four times better.

Barbara Romanowicz: Right. I'm talking about the global resolution, if you're looking at a global model of the Earth, because on the regional scale, if you have many, many stations and also many earthquakes, you can obtain much higher resolution. But this would be for limited regions, generally just in the upper 200 or 300 kilometers of the mantle. You can obtain resolutions of several tens of kilometers. The smaller the region you look at, the higher the frequency of the waves you want to use, all the way to what is done in exploration geophysics, where they are looking at very, very local kinds of structures, with very high, sub-kilometer resolution.

Oliver Strimpel: And does that enable us, at that fine scale, also to map the plumbing underneath volcanoes?

Barbara Romanowicz: So the plumbing underneath volcanoes, and in particular the volcanoes at hotspots: the problem there is that unless you have arrays of ocean-bottom seismometers, we don't have enough so-called aperture to be able to go very deep. There have been attempts in Iceland and in Hawaii, and also Japanese scientists have put some ocean-bottom stations in the southwest Pacific, to try and improve the resolution deeper and deeper. However, it is still limited, especially if you only use one type of wave, like only the P wave or the S wave, because you don't have enough illumination from different directions, deep enough, to see the whole column.

Oliver Strimpel: Changing tack a little bit, I've heard of a recent development called ambient noise tomography, in which we use background seismic noise generated not from earthquakes at all, but from the action of things like ocean waves smashing against rocks. What can that tell us about the Earth's interior?

Barbara Romanowicz: So traditionally, we use the signal generated by natural earthquakes, which informs us on the structure that the waves have traversed. However, it turns out that in the absence of earthquakes, there is also signal that can be interpreted in terms of arriving seismic waves. This was discovered 20 years ago by seismologists in Japan who were looking at the spectra of the Earth's normal modes. If you wait long enough, the surface waves that travel around the Earth will interfere with each other to create standing modes, and the frequencies of the standing modes inform you on the internal structure of the Earth. That's another kind of observable that we use. We call them spectra. We transform the time-domain recording, over a long period of time, like 24 hours, into a spectrum, that is, into the frequency domain, where we see these peaks, which are characteristic of the Earth's structure. And we can do this after large earthquakes that generate enough energy for the waves to circle around the Earth many times. But these Japanese seismologists discovered that you could actually see these peaks in the spectra even when there are no earthquakes. That is, they eliminated all the time intervals where there were earthquakes, and then they stacked all the remaining records, and they still saw the peaks characteristic of the Earth. And this started the field of looking at the Earth's hum. And it turns out that the noise that is the source of these seismic waves is primarily located in the oceans. It's the result of interaction of the wind during strong storms in the ocean with the ocean waves, and then interaction of ocean waves with the ocean floor, that gives rise to sources of seismic energy. That's just a curiosity, in a way, although it is now being used more and more to also study the Earth's deeper structure. But there is another frequency band, which is called the microseismic band, roughly between 3 seconds and 15 seconds of period, where there is always a lot of noise, much, much stronger noise than everywhere else in the recordable frequency band. And these waves also originate in the oceans. They are also related to interactions of wind with waves and with the seafloor, but they're more related to actual waves breaking on the shore, and more in line with where the strongest energy is in the oceans. And methodologies have been developed in the last 15 years to use this noise to extract the propagation properties of, primarily, the surface waves, by a technique which comes from the fact that when you have a uniform distribution of sources around two stations, if you cross-correlate the recordings at these two stations in the absence of earthquakes, what will come out is the wave field, the record that you would observe if one of the stations was the source of the waves and the other one was an observing station. And so this has given rise to a methodology called ambient noise tomography, which uses this very strong noise in this period band, around 10 to 30 seconds, to infer the structure of the Earth's crust, and that has been very effective, because now all you need is a good network of stations. You don't need earthquakes. All you need is a good distribution of stations, so that you can measure these properties between pairs of stations and get a good image of the structure beneath this ensemble of stations.

Oliver Strimpel: That's amazing. No earthquakes. You just have the ambient noise that just keeps exciting the Earth's normal modes.

Barbara Romanowicz: That's right.

Oliver Strimpel: If I were to write you a blank check, and you could have as many resources as you wanted, what would you use it for?

Barbara Romanowicz: Well, apart from having access to supercomputers as much as we need, the other aspect is to try and install stations on the ocean floor, so that we can complement the distribution of stations that we have on land and on the very few islands that exist around the oceans.

Oliver Strimpel: Barbara Romanowicz, thank you very much.

Barbara Romanowicz: Thank you.

Oliver Strimpel: For more about Geology Bites, as well as pictures and diagrams that illustrate this podcast, you can go to geologybites.com.