** Next:** Synthetic Electroseismic data
** Up:** Haines and Guitton: Electroseismics
** Previous:** Haines and Guitton: Electroseismics

Electroseismic phenomena produce two forms of energy: the interface
response signal, and the coseismic noise Haines and Guitton (2002). Electroseismic data
processing must attenuate the coseismic noise in order to reveal the
interface response. In short, the
signal is composed of flat events (the interface response has virtually zero moveout) while the
noise (coseismic energy) has moveout similar to seismic data. The
object is to remove the curved energy so that it does not contaminate
the final stack of the gather. We
employ the signal/noise separation technique described by
Guitton (2003) to test its effectiveness in electroseismic
processing. The basic approach is to estimate non-stationary
prediction-error filters (PEF's) for the signal and the noise, and to
use these PEF's in an iterative signal/noise separation following Guitton et al. (2001).
The electroseismic signal is far weaker than
the noise so we can not hope to obtain an adequate model for the
estimation of signal PEF's by muting in the parabolic radon transform (PRT) domain, or other
alterations of the original data. We can, however, take advantage of the
fact that the amplitude pattern of the signal can be easily modeled.
It is the pattern of the potential (*V*) of a dipole field:

| |
(1) |

as measured at a horizontal offset *x* from a dipole at depth *z*, where
*q* is the magnitude of the electrical charges, *d* is the distance between the two
separated charges, and is the electrical permittivity of the
Earth.
Using a velocity model to provide the relationship between depth and travel
time, and making basic assumptions about the size of the Fresnel zone
producing the dipole, we can compute a model of the relative amplitude
measured at various locations for events corresponding with any given
travel time. This amplitude pattern may be used directly in the
estimation of one-dimensional (in the offset direction) PEF's (since
such a one-dimensional PEF contains no wavelet information), or
we can use the amplitude pattern to scale synthetic wavelets to be
used as models for PEF estimation. Thus if we have a velocity model for a particular
study area, we can estimate non-stationary signal PEF's to target
any interface response events that may be in the data without the need
for *a priori* knowledge of their arrival times.
We use simple physics [equation
(1)] to design general signal PEF's, and use components
of the original data to design noise PEF's. We show that these
non-stationary PEF's provide an effective means for
separating the interface response signal from the coseismic
noise in synthetic and real electroseismic data.

** Next:** Synthetic Electroseismic data
** Up:** Haines and Guitton: Electroseismics
** Previous:** Haines and Guitton: Electroseismics
Stanford Exploration Project

7/8/2003