- © 2009 by the Seismological Society of America
Realistic 3D models of the seismic velocity distribution within the Earth have the potential to significantly improve our ability to accurately and precisely locate seismic events around the world (e.g., Flanagan et al. 2007). Development of these models by tomographic inversion of observed seismic travel times and the use of such models in seismic event location calculations, require the ability to accurately and efficiently compute predicted source to receiver travel times through 3D velocity structures. Meeting these requirements on a global scale presents two fundamental challenges: 1) how to represent the 3D distribution of seismic velocity in the Earth in models that can be interpolated efficiently, and 2) how to efficiently and accurately compute predicted travel times through these models. We present a 3D variable-resolution modeling methodology and a travel time prediction algorithm based on ray theory that address these issues and are compatible with computational resources ranging from laptop computers to powerful distributed computing systems.
EARTH MODEL REPRESENTATION
A simple representation of the 3D distribution of Earth properties is a regular latitude-longitude grid over the surface of the Earth with grid nodes distributed in an irregular manner in the radial direction (Figure 1A). Finding the grid cell in which an arbitrary point is located in such a model is both straight-forward and efficient. For models that do not extend to polar regions (e.g., Pasyanos et al. 2004), this approach provides satisfactory performance, but for global models such as the Crust 2.0 model (Bassin et al. 2000; http://igppweb.ucsd.edu/~gabi/crust2.html) the approach suffers from the fact that meridians intersect at the poles, leading to much smaller grid cells at high latitudes than at the equator, when cell size is evaluated in km2.
Over the past couple of decades gridless Earth model representations and several representations involving irregular meshes have been …