When a 7.9-magnitude earthquake struck central China's Sichuan province in 2008, seismic waves rippled through the region, toppling apartment houses in the city of Chengdu and swaying office buildings 1,000 miles away in Shanghai.
Though destructive, earthquakes provide benefit in one respect: they help researchers learn about the structure of the Earth, which in turn could lead to more accurate predictions of damage from future quakes and volcanic activity. By eavesdropping on the seismic vibrations of quakes as they rumble through the Earth, researchers can detect the existence of structures such as mineral deposits, subterranean lakes, and upwellings of magma. Thanks to a growing earthquake detection network and superfast computers, geoscientists are now able to explore the Earth's interior, a region that has been more inaccessible than the deepest ocean or the farthest planet in our solar system.
Princeton geosciences professor Jeroen Tromp and his team have embarked on an ambitious project to use earthquakes to map the Earth's entire mantle, the semisolid rock that stretches to a depth of 1,800 miles, about halfway down to the planet's center and about 300 times deeper than humans have drilled. For the task, his team will use one of the world's fastest supercomputers, Titan, which can perform more than 20 quadrillion calculations per second and is located at the Department of Energy's Oak Ridge National Laboratory in Tennessee.
"Seismology is changing at a fundamental level due to advances in computing power," said Tromp, who earned his Ph.D. in geology from Princeton and is Princeton's Blair Professor of Geology, professor of applied and computational mathematics, and associate director of the Princeton Institute for Computational Science and Engineering. "If someone had told me what seismology would look like 20 years from when I graduated from Princeton in 1992, I would have never believed it."
For the project, Tromp will use seismic waves from roughly 3,000 quakes of magnitude 5.5 and greater, recorded at thousands of seismographic stations worldwide and distributed via the National Science Foundation's Incorporated Research Institutions for Seismology. These stations make recordings, or seismograms, that detail the movement produced by seismic waves, which typically travel at speeds of several miles per second and last several minutes.
"The ultimate goal is a 3-D map on a global scale," said Tromp, who expects to have preliminary results at the end of this year. "We are specifically interested in the structure of mantle upwellings and plumes," he said, "but much of it will be investigating the images for unusual features."
These unusual features could include, for example, a fragment of a tectonic plate that broke off and sank into the mantle. The resulting map could tell seismologists more about the precise locations of underlying tectonic plates, which can trigger earthquakes when they shift or slide against each other. The maps could also reveal the locations of magma that, if it comes to the surface, causes volcanic activity.
As seismic waves travel, they slow or speed up depending on the density, temperature and type of rock. For example, they slow down when traveling through an underground aquifer or magma. By combining seismograms from many earthquakes recorded at many stations, geologists can produce a three-dimensional model of the structure under the Earth's surface.
This technique is called seismic tomography and is analogous to computerized tomography used in medical ("CAT") scans, in which a scanner captures a series of X-ray images from different viewpoints, creating cross-sectional images that can be combined into 3-D images.
Over the past eight years, Tromp has been at the forefront of research on how to improve seismic tomography to obtain high-definition, accurate images of the Earth's interior. Past approaches for making these images incorporate only three types of seismic waves: primary or compressional waves, secondary or shear waves, and surface waves. Tromp has pioneered techniques for using much more of the information in seismograms, utilizing both waves that travel from the quake epicenter to the detector as well as those that travel from the detector to the quake, which are called adjoint waves.
"If we are going to do better, we need to use everything in the picture, in other words, everything and anything in the seismograms," Tromp said.
Tromp's team feeds data from seismograms into a computer model, which simulates each wave as it propagates from the epicenter. The resulting "synthetic seismograms" are compared to real seismograms, and the differences are fed back into the model to improve it. The researchers do this over and over, comparing data with simulations and extracting differences. With each pass, they improve their model.
Before attempting to map the entire Earth, Tromp and his team, with funding from the National Science Foundation, showed that their technique worked in smaller regions, first in Southern California using data from 143 quakes, which was published in the journal Science in 2009, and later in Europe using 190 quakes. The Europe study, published in the journals Nature Geosciences in 2012 and Science in 2013, included work by then graduate student Hejun Zhu, now a postdoctoral researcher at the University of Texas-Austin; postdoctoral researcher Ebru Bozdağ, now an assistant professor at the University of Nice Sophia Antipolis; postdoctoral researcher Daniel Peter, who today is a senior scientist at ETH Zurich; and David Pugmire, a visualization scientist at the Oak Ridge Leadership Computing Facility (OLCF).
"Multi-parameter adjoint tomography of the crust and upper mantle beneath East Asia – Part I: Model construction and comparisons" Journal of Geophysical Research: Solid Earth. DOI: 10.1002/2014JB011638