3D Tectonic Modeling on Supercomputer Answers Old Geology Question

The theory of plate tectonics predicts that the outer layer of the earth is composed of approximately 15 tectonic plates that are in motion with respect to one another, and that the deformation of those plates will be concentrated at the boundary – where plates meet. One of the most interesting things about the central Alaska Range, and Mt. McKinely in particular, is that it's located nearly 500 kilometers away from the boundary. 

Alaska is an example of a subduction zone where two tectonic plates, the Pacific and North American, move towards each other with the denser Pacific plate descending beneath the overriding North American plate. “There are a lot of different angles that a plate can descend,” says Margarete Jadamec, a US National Science Foundation postdoctoral fellow in the Department of Geological Sciences at Brown University in Providence, Rhode Island, US. “But our work shows that both flat slab subduction and the Denali fault are required to form the intercontinental mountain range." Jadamec’s findings are published in Earth and Planetary Science Letters.

For decades, tectonic modeling has embraced a limited, two-dimensional paradigm– imagine drawing one horizontal line on paper, and another line that descends at a small, constant angle below the first (flat slab subduction). However, Jadamec and scientists at the University of California, Davis, US, are exploring 3D tectonic modeling and pushing the limits of high-performance computing.

“I included a shallow dipping slab and found that the shallow dip wasn't enough to create the central Alaska Range. I needed to include another first order feature – the Denali fault, which is also 500 kilometers away from the plate boundary,” says Jadamec. “The shear zone I’ve added to the numerical models represents the location of this fault.”

“We’ve also integrated an entire suite of geologic and geophysical observations collected by geologists and seismologists,” Jadamec says. This enabled Jadamec, and colleagues Magali Billen and Sarah Roeske, to create a very detailed tectonic representation of a plate boundary, including a detailed temperature structure, density structure, and viscosity structure.

Jadamec is looking at the formation of the central Alaska Range over a million-year time scale. “The code I’ve customized and use, called CitcomCU, is just one of many community supported codes that the Computational Infrastructure for Geodynamics (CIG) develops and supports for the earth science community,” says Jadamec.

The Earth behaves as a viscoelastic fluid, responding elastically on short timescales from seconds to days. But it behaves viscously over larger time scales of millions of years. “CitcomCU solves the basic conservation of mass momentum and energy for the viscous flow. But what’s different from computational fluid dynamics experiments in other areas of science is that the viscosity of the earth is extremely large; ranging from 10¹⁷ to 10²⁴ Pascal-seconds in the million-year time window we work with.”

A Pascal-second can be thought of as thickness or internal friction. In this window of extremely high viscosity, the Earth’s flow rates are very small, ranging from millimeters to tens of centimeters per year.

For the fourteen 3D geodynamic models of the Alaska subduction zone Jadamec developed, the viscosity structure varied up to seven to eight orders of magnitude over 100 kilometers within the model. “To resolve the tectonic questions I was interested in, you have to go to that level of detail, which is quite a challenge for the numerical solvers,” explains Jadamec.

Each model ran for approximately 17,000 compute hours. The finite element mesh of 960 × 648 × 160 elements included nearly 100.1 million nodes. Compute time was available via an XSEDE allocation on the Lonestar cluster at the Texas Advanced Computing Center (TACC) at The University of Texas at Austin, US.“The 3D visualization was done on my in-office 3D virtual reality station at Brown University, and in collaboration with Oliver Kreylos and Burak Yikilmaz at the UC Davis KeckCAVES.

“When I'm modeling an inherently three-dimensional object like the Earth, that is represented by hundreds of millions of numbers, being able to project those numbers into a virtual space is a critical,” Jadamec says. Her in-office 3D virtual reality station is an important piece of her workflow.

“For my work with larger data sets, I use a Linux box, a flat screen HD 3D TV, and a Razor game controller. You can configure open source visualization software to the dimensions of the screen and how it projects images. But the cool thing is that it’s portable. You can plug a laptop in just as easily for use with smaller data sets.” 

Jadamec uses the virtual reality station in her undergraduate tectonics lab at Brown. “Once you put on the glasses, as your model is projected – the Earth for example – you can manipulate and move the object using the Razor game controller.” The interactive and immersive environment makes it easier for students to conceptualize large spatial scales. “It is really a different teaching experience. I actually have students that thank me. It helps make geology real.”

Jadamec is now turning her attention to subduction zones that have multiple subduction plates, as well as links to the chemistry of volcanoes in Central America in collaboration with seismologist Karen Fischer at Brown University. “Previous numerical models were really very simple. But now with advances in computational infrastructure they have really taken off. We’re now able to really push the limits of the scientific questions we can ask.” -- by Amber Harmon, © i SGTW