Octree-based finite element method for large-scale earthquake ground motion modeling in heterogeneous basins

This work reports on the development of a parallel numerical octree-based finite element methodology for simulating large-scale earthquake-induced ground motion in highly heterogeneous basins. We target large sedimentary basins with contrasts in wavelengths of over an order of magnitude. We overcome the problem of multiple physical scales by using finite elements on locally-resolved hexahedra derived from octree-based grids. The extremely large mesh sizes require special mesh generation techniques. We use an out-of-core octree-based mesh generator developed by Tu and O'Hallaron (2002), which allows us to generate meshes of essentially arbitrary resolution. Despite the method's multi-resolution capability, large problem sizes necessitate the use of distributed memory parallel supercomputers to solve the elastic wave propagation problem. We have developed a system that helps automate the task of writing efficient portable octree-based mesh solvers for distributed memory parallel supercomputers. The numerical methodology and software system have been used to simulate the seismic response of both idealized systems and of the Greater Los Angeles basin to simple pulses and to an aftershock of the 1994 Northridge Earthquake, for frequencies of up to 1 Hz. We report on parallel performance on the Cray T3E and the Terascale Computing System (TCS) at the Pittsburgh Supercomputing Center for several models, ranging in size from 40,000 to 200 million hexahedra. The results indicate that, despite the highly irregular structure of the problem, excellent performance and scalability are achieved.