Thursday 12 May 2016

A collaborative research project between the University of Sydney and the Californian Institute of Technology has solved a long standing geological mystery – just how the distinct bend in the Hawaiian-Emperor Seamount Chain came to be.

Using NCI's Raijin supercomputer, the research team has been simulating flow patterns in the Earth's mantle over the past 100 million years. By studying these flow patterns, the team recognised that the changing formation of the Hawaiian-Emperor Seamount Chain was ultimately caused by motion close to the Earth's core.

PhD candidate Rakib Hassan has led the team's efforts to debunk the previously held hypothesis - that the chain's unique bend was largely the result of plate motion and shallow mantle activity.

"These findings suggest the shape of volcanic seamount chains record motion in the deepest mantle, near the Earth's core. The more coherent and rapid the motion deep in the mantle, the more acute its effects are on the shape of seamount chains above," said Mr Hassan in today's media release.

 The paper, 'A rapid burst in hotspot motion through the interaction of tectonics and deep mantle flow' was published today in Nature.

"Our results help resolve a major enigma of why volcanic seamount chains on the same tectonic plate can have very different shapes," said Professor Dietmar Müller from the University's EarthByte Group.

"It is now clear that we first need to understand the dynamics of the deepest 'Underworld', right above the core, to unravel the history of volcanism at the Earth's surface."

According to Mr Hassan, the allocation of computing capacity and storage space on Raijin allowed their research to be fast-tracked.

"Spherical models of convection within the Earth's mantle are computationally expensive and the cost increases dramatically with increasing resolving power – leading to increased runtimes that can range from weeks to even months," explains Mr Hassan.

"Since the main thrust of this project focused on modelling mantle plumes – around 100–200 kms in diameter – in a convecting mantle, a high resolving power of the models was a prerequisite at the outset. We have been able to keep the runtimes of these models manageable – three to four days per model – by distributing the computations over many hundreds of CPU cores on Raijin."

"Over the course of the project, we ran over a hundred models that produced tens of terabytes of data that had to be transferred from Raijin, post-processed and analysed to glean valuable insights.

The University of Sydney is currently using Raijin to run previously impossible computational models to gain an understanding of how continents and sedimentary basins are formed. The ARC Basin GENESIS Hub forms the backbone of this geologic research.

Associate Professor Patrice Rey says the Hub's purpose is "to understand the formation of continental margins and sedimentary basins in the context of the global Earth, in which we account for the mantle flow, tectonic processes and surface processes."

The team is working with 4 different models, each looking at a different element of the sedimentary basin process, including mantle flow, plate tectonics and surface effects.

"Even better", Professor Müller adds, "all software developed in the Basin Genesis Hub is open-source.  This drives greater innovation because if you can get other smart people to care about what you are doing, contributing to what you're working on, you'll get a lot more done. While most of the coding in the Hub is done by a small number of people, the collective intelligence of a larger group, adding their personal experience and feedback, is what makes open source software so powerful. I believe that we have the first industry hub that is firmly built on these insights".

The Basin Hub is progressively working on coupling the individual models, so that all the important factors of the Earth's geodynamics are being taken into account in one all-encompassing model. Such a model is still 4 to 6 years away but along the way there are exciting discoveries to be made. Researchers have been able to model the evolution of the Australian landscape over the past 150 million years, using the Badlands model that they developed to look at erosion and sediment transportation.

"The capacity of the code to simulate the evolution of a landscape, that's really mindboggling" says Associate Professor Rey, "That has definitely attracted a lot of attention from our colleagues and industry partners here. I don't think there is any place in the world where this has been done before. So it's definitively a first."

To run all of these models, the group has been awarded 11 million computing hours on Raijin and Magnus for 2016, through the National Computational Merit Allocation Scheme, which was one of the top 4 allocations across the country. Associate Professor Rey says, "We are really excited about the future and we are really grateful to have access to such a cutting-edge infrastructure here in Australia."

"This is a very exciting time in science," he explains, "The equations we solve are the Navier-Stokes equations. Those equations have been with us for 150 years, but it is only right now that we can really use those mathematical concepts developed a century and a half ago, combine them with computer science from the mid twentieth century and now we have the infrastructure, the computational power to make use of all these developments, and that's really exciting."

The research that the Basin GENESIS Hub team is driving relies on the High Performance Computing that NCI and the other HPC facilities around the country provide. "We are not aware of any other infrastructure that can accommodate our numerical modelling. That kind of computational power is out of reach of most universities so we rely on national infrastructure for this research to transform our understanding of the structure and evolution of sedimentary basins."