Dept. of Geosciences
Princeton University
Princeton
NJ 08544
Email:
briansch@princeton.edu
poster/oral: poster
| Global seismology is at a critical stage in its maturation as a field of study. With the advent of broadband digital seismic networks, we have been inundated with a wealth of high-quality data. This new data has brought global traveltime tomographic images close to their saturation point, i.e., the point at which additional measurements make a negligible difference in the major features observed, given present station coverage and the natural source distribution. More stringent seismic constraints on the details of the Earth's deep structure can be obtained only by including the additional information recorded in the seismic waveform. This valuable information is most effectively extracted by the comparison of data with synthetic waveforms computed for a proposed structure model. The need for a practical and accurate method of generating these synthetics for the propagation of broadband seismic waves through realistic, 3-D models has never been greater. However, at the present time practicality and accuracy are almost mutually exclusive. Practicality is achievable if one is willing to settle for high-frequency asymptotic solutions or for models with merely 1-D variations or for both. Until recent years, these "practical" methods were really the only methods available to seismologists, but now those who have plenty of computing power and memory, and plenty of patience, can generate very accurate seismograms up to frequencies that are approaching the range of low-frequency teleseismic S-waves. Here we present an alternative approach, one that provides an adjustable compromise between computational speed and accuracy. We recast an existing path integral solution for acoustic wave propagation in a form that is immensely more amenable to numerical implementation. Several comparisons between the new path integral method and slower, established numerical methods for nontrivial 3-D media will be made. |