HR: 16:30h
AN: S12G-03 [PDF]
TI: Estimation of two dimensional von Karman stochastic parameters from the seismic coda
AU: * Poppeliers, C
EM: poppelie@caam.rice.edu
AF: Rice Center for Computational Geophysics, 6100 Main Street
Rice University Mail Slot 126
P.O. Box 1892, Houston, TX 77005 United States
AU: Levander, A
EM: alan@rice.edu
AF: Rice Department of Earth Science, 6100 Main Street
Rice University Mail Slot 126
P.O. Box 1892, Houston, TX 77005 United States
AU: Symes, W W
EM: symes@caam.rice.edu
AF: Rice Department of Computational and Applied Mathematics, 6100 Main Street
Rice University Mail Slot 134, Houston, TX 77005 United States
AB:
Coda waves are generated from singly and multiply-
scattered primary waves in the Earth. The scattered
wavefield is composed of converted phases, transmissions,
and reflections arising from three dimensional fluctuations in seismic velocities and density. Statistical models of the
complex medium are customarily used to explain seismic scattering. Our goal is to
invert the coda directly for stochastic properties of the medium through which the seismic energy travels. We characterize
the medium
with the von Karman autocorrelation
model which can accurately model the statistical distribution of
velocity/density variations of the
crystalline portions of the
Earth's crust. Synthetic data generated through von Karman media closely resemble real data. The von Karman
autocorrelation model mimics the Earth's self affine nature
and can be described in two-dimensions by three parameters; 1) the horizontal characteristic length, 2) the vertical
characteristic length,
and 3) the Hurst exponent, which describes the roughness of
the statistical model. We present two inversion schemes to estimate von Karman parameters. In the first scheme, we show
that inverting seismic array data for the lateral characteristic length is possible, but that the inversion has to be
calibrated
against the central frequency of the input seismic wavelet.
In the second inversion scheme, we make use of the multiplicative
relation between the power spectra of the observations and that of the source pulse and the stochastic medium. Thus, if we
have a good estimate
of the seismic pulse, we can invert directly for von Karman
parameters. In the absence of complete knowledge of the
pulse, we can formulate a nonlinear inversion and
estimate the seismic pulse and the von Karman parameters simultaneously.
The products of our inversions are cross sections of
stochastic properties that can define different areas of
texture and/or fabric in the Earth. We demonstrate our
inversion on synthetic data as well as a seismic reflection
section from the CDROM experiment in the western U.S. With
two dimensional, synthetic seismic reflection data, we
show that we can recover the dip-dependant
lateral characteristic length. Additionally, we present
results of inverting for the lateral characteristic length
using the CDROM seismic reflection dataset.
We claim that inverting for lateral characteristic length can be a
useful tool to place bounds on the base of the crust, which
is difficult to interpret deterministically in the processed seisic line.
Although preliminary results of inverting for the vertical
characteristic length with synthetic data are promising, we
are still in the process of refining the process. Namely,
we find that the primary limitation of the inversion is
accurate information of the source pulse. However, with a
reasonable estimate of the source pulse power spectrum
non-linear inversion holds promise.
DE: 0902 Computational methods, seismic
DE: 0935 Seismic methods (3025)
DE: 7203 Body wave propagation
DE: 7294 Instruments and techniques
SC: Seismology [S]
MN: 2003 Fall Meeting