Beyond traditional processors in geosciences

On average, 1 100 earthquakes with potential to cause injury and loss of life, as well as damage to built infrastructure, occur every year. The design of a human environment with appropriate earthquake resistance is, however, a major scientific and engineering challenge. The main difficulty is the complexity of physical processes in the Earth's subsurface.

Much has been learned about fault rupture processes through the scientific study of measurements of ground motions. Drawing upon the advancements in computational infrastructure, there is a timely opportunity for accurate numerical simulations. Physics-based models offer a better understanding of the way that subsurface geologic structures influence ground motions.

The aim of the EU-funded project HPC-GA (High performance computing for geophysics applications) was to exploit and tune the existing computational infrastructure within European facilities to earthquake hazard assessment. For this purpose, an international, multidisciplinary consortium of European and South American researchers was gathered.

Team members evaluated the functionalities provided by current runtime systems and pointed out their constraints. Simulations of earthquake ground motions need to cover several subsystems operating at different scales. Modelling the coupling of subsystems is essential, but is limited by the computational power required to handle an increasing amount of data.

The next step was, therefore, to develop new methods for efficient scheduling of processes and clever data distribution on multicore, shared-memory computational environments. In addition, they implemented data-adaptive beam-forming algorithms on a parallel computing architecture to obtain useful information from even noisy seismic wave data.

Dedicated algorithms were developed for the 3D joint inversion of seismic, magnetotelluric and gravity data. Data acquired in geological surveys is usually restricted to the Earth's surface or the shallow subsurface, often with large spacing between measurement sites. The joint approach pursued promises models that can explain all data simultaneously.

The HPC-GA project has resulted in many advances in computational technology and numerical methods to carry forward models of geodynamics and seismic wave propagation at unprecedented resolution. More reliable information will lead to a deeper understanding of the physics of earthquakes and a more accurate estimation of the seismic risk for critical infrastructure.