S33C-2785
Spectral-Element Seismic Wave Propagation Codes for both Forward Modeling in Complex Media and Adjoint Tomography

Wednesday, 16 December 2015
Poster Hall (Moscone South)
James A Smith1, Daniel B Peter2, Jeroen Tromp1, Dimitri Komatitsch3, Matthieu Philippe Lefebvre1 and SPECFEM3D Development Team, (1)Princeton University, Princeton, NJ, United States, (2)King Abdullah University of Science and Technology, Physical Sciences and Engineering, Thuwal, Saudi Arabia, (3)Université de Pau, Pau Cedex, France
Abstract:
We present both SPECFEM3D_Cartesian and SPECFEM3D_GLOBE open-source codes, representing high-performance numerical wave solvers simulating seismic wave propagation for local-, regional-, and global-scale application. These codes are suitable for both forward propagation in complex media and tomographic imaging. Both solvers compute highly accurate seismic wave fields using the continuous Galerkin spectral-element method on unstructured meshes. Lateral variations in compressional- and shear-wave speeds, density, as well as 3D attenuation Q models, topography and fluid-solid coupling are all readily included in both codes. For global simulations, effects due to rotation, ellipticity, the oceans, 3D crustal models, and self-gravitation are additionally included. Both packages provide forward and adjoint functionality suitable for adjoint tomography on high-performance computing architectures.

We highlight the most recent release of the global version which includes improved performance, simultaneous MPI runs, OpenCL and CUDA support via an automatic source-to-source transformation library (BOAST), parallel I/O readers and writers for databases using ADIOS and seismograms using the recently developed Adaptable Seismic Data Format (ASDF) with built-in provenance. This makes our spectral-element solvers current state-of-the-art, open-source community codes for high-performance seismic wave propagation on arbitrarily complex 3D models. Together with these solvers, we provide full-waveform inversion tools to image the Earth's interior at unprecedented resolution.