The quality of tomographic images of the Earth’s interior and earthquake source models is
closely tied to our ability to efficiently and accurately simulate 3D seismic wave propagation.
For decades seismologists have used asymptotic, approximate methods to address the
forward problem in seismology, namely, given a seismic source and a 3D Earth
model, accurately simulate the associated wave motions. In recent years, modern
numerical methods and parallel computers have facilitated fully 3D simulations of
seismic wave propagation at unprecedented resolution and accuracy, heralding the
age of computational seismology. The current focus is on harnessing the power of
these sophisticated forward modeling tools to enhance the quality of images of the
Earth’s interior and the earthquake rupture process, that is, to address the inverse
problem.
Traditional tomographic methods utilize traveltime and dispersion information obtained
by comparing data with simulations, and interpret such measurements based on ray theory or
other approximate methods. Because of the limitations of these approximate techniques, only
certain parts of seismograms can be used, and initial models are generally restricted to be
layered or spherically symmetric. With modern numerical modeling tools we are now
going well beyond classical tomography, using fully 3D initial models and utilizing
as much information contained in seismograms as possible. The ultimate goal is
broad band full waveform inversion utilizing entire seismograms. Surprisingly, one
tomographic iteration may be performed based on just two numerical simulations
for each earthquake: one calculation for the current model and a second ‘adjoint’
calculation that uses time-reversed signals at the receivers as simultaneous, fictitious
sources.
Seismic imaging based on adjoint methods assimilates seismographic information into 3D
models of elastic (seismic wavespeeds) and anelastic (quality factors) structure. These
methods fully account for the physics of wave excitation, propagation, and interaction by
numerically solving the inhomogeneous equations of motion for a heterogeneous anelastic
solid. Such methods require the execution of complex workflows, involving extensive pre-
and post-processing of observed and simulated seismograms. After successful applications of
adjoint tomography in southern California and Europe, we are currently moving toward
adjoint tomography of the entire planet. The objective is to image Earth’s global interior
based on full waveform inversion, thereby facilitating a deeper understanding of its
physics and chemistry. To start with, we selected 255 earthquakes and gathered data
from global seismographic networks. Our strategy is to invert for crust and mantle
structure jointly, thereby avoiding any bias introduced into upper-mantle images
due to commonly used ‘crustal corrections’. Our ultimate aim is to harness more
than 6,000 magnitude 5-7 earthquakes digitally recorded over the past 20 years. |