What do seismic waves really resolve in an isotropic 3D Earth?
Guust Nolet, Henk Marquering, Tony Dahlen, Sergei Lebedev, Jun
Tong (1)
and Thomas Meier (2)
(1) Department of Geosciences, Princeton University
(2) Friedrich
Schiller Universität, Jena
poster/oral: poster
On regional scales, there is often disagreement between tomographic
models. Noise in the seismograms is probably a minor factor in
causing such differences (although phase velocity measurements at
low frequencies may be affected by atmospheric disturbances). More
important, in our view, are the errors introduced by various
assumptions made in the interpretation process, such as:
- The assumption that ray theory is valid,
- The assumption that interfering (higher) mode energy can be
windowed
out when measuring phase or group velocity of Love or Rayleigh
waves
- The assumption that scattered wave energy (or mode-mode
conversions)
can be neglected.
The interference of higher mode energy is no problem if waveforms
are fitted, and no attempt is made to isolate a particular mode.
However, waveform tomography currently disregards the interaction
with higher modes due to scattering.
We have used Born theory for the scattering of surface wave modes
(Snieder and Nolet, J. Geophys., 61:55-63,1987) to analyze different
aspects of mode-mode conversions. Using a surface wave formalism
for mode coupling greatly increases the efficiency of such
calculations
with respect to methods using discrete modes, and this enables us
to calculate Frechet kernels in 3D. We show how one seismogram, which
shows severe Love wave energy on the vertical component, can be
fitted with 100% variance reduction assuming a 3D model of reasonable
scattering strength. In a synthetic test, we show how scattered waves
allow us to reconstruct a point scatterer using only one seismogram
(Meier et al, JGR 1997, in press). We are currently implementing
the theory in a new method of diffraction tomography for S waves.
We also show how the theory can be applied to compute 3D Frechet
kernels for body waves. We do this for waveforms as well as for
travel time anomalies as determined by crosscorrelation. The
resulting kernels often defy intuition. For example, under certain
conditions the sensitivity of the Earth to the travel time of a
body wave is minimal on the ray itself, and greatest on a tube around
the ray. Minimax phases such as SS have very complicated Frechet
kernels. Conversions between the fundamental mode and higher modes
induce a crustal sensitivity to body waves over large segments of the
wavepath, remote from the ray location.
Finally, we show results from finite difference calculations. Large
differences may exist between 'actual' travel times (from the FD
simulations), ray-theoretical times determined with or without
the Fermat approximation, and travel times as predicted by
first-order
Born theory. Again results are sometimes unexpected: we shall show
an example where a negative velocity anomaly induces a faster arrival
time as determined by crosscorrelation, and offer an explanation in
terms of higher order Fresnel zones.
These results are very relevant when discussing a new reference Earth
model. What started as a theoretical curiosity, is quickly turning
out to become a major investigation into the validity of many
assumptions
that are commonly accepted to be true, and that underlie most if not
all
of the seismological imaging of the deep Earth. We conjecture that
errors in theoretical assumptions are currently the major hurdle to
get
agreement about the extent and shape of 3D heterogeneity in the
Earth.
Guust Nolet (
guust@geo.princeton.edu)
Go back to
List of Participants
Go to
main page