In a study recently published in the Journal of Geophysical Research, researchers Saule Simute and Andreas Fichtner used supercomputers to identify a reservoir of magma starting about 50 km below Ulleung, a small island in the Sea of Japan. Dubbed the Ulleung anomaly, the magma reservoir is about 300 km wide and 200 km deep. In contrast, the island is only about 10 km wide.

Computer models

To locate the lava, Fichtner’s team started with seismic wave records of 58 earthquakes, for a total of 5,500 three-component waveform data sets. They then compared the waveforms with seismic wave simulations, optimized through 19 iterations to ensure the computer models matched observations from the seismic stations. These simulations produced a three-dimensional rendering of the crust and upper mantle of the Japanese islands and surrounding area.

Enormous computational requirements

“What is new about our method is its ability to extract as much information as possible from the data – and this is why we manage to discover new aspects of earth structure, such as this reservoir,” Fichtner, professor at the ETH Institute of Geophysics, told Science Node. “However, the price to pay is enormous computational requirements.”

Fichtner’s simulations clocked about 10 million CPU and GPU hours on Piz Daint (see image above) at the Swiss National Supercomputing Center (CSCS) and the now decommissioned Huygens, the national supercomputing center of the Netherlands operated by SURFsara. SURFsara is one of the operating companies of the Dutch research and education network SURF.

Published: 09/2016

For more information please contact the contributor/s: