next up previous print clean
Next: Conclusions and future work Up: Haines and Guitton: Electroseismics Previous: Synthetic Electroseismic data

Real Electroseismic data

We next test non-stationary PEF's on real data (Figure [*]a). This 48-channel shot gather was constructed by interleaving two 24-channel gathers that were collected according to the methods described by Haines et al. (2003) and Haines and Guitton (2002). The shot point is the same for the two records, and the receiver positions are shifted by half the receiver spacing for the second record. Thus the resulting 48-channel record has half the original receiver spacing, and provides a test gather that can be processed more easily than the original 24 channel records. The shot point is in the center of the dipole receiver line, in the geometry that would be used for a typical survey. The gather shown in Figure [*]a has been pre-processed to remove 60 Hz energy and eliminate high- and low- frequency background noise.

 
NS_r1fig
NS_r1fig
Figure 4
a) Field data after pre-processing. b) Result after signal/noise separation. Note that coseismic energy is almost entirely removed. c) Model used for estimation of signal PEF, based on equation (1). d) Muted version of field data, used as model used for estimation of noise PEF.
[*] view burn build edit restore

Using the coseismic energy as a guide, we establish a basic velocity function to be used with equation (1) for the determination of the model to be used for signal PEF estimation (Figure [*]c). For estimation of the noise PEF we use a windowed version of the original data file, shown in Figure [*]d. This choice of noise model is effective, as shown by the final result (Figure [*]b), but is somewhat unrealistic, as we would ideally be looking for signal within the noisy part of the data and would not want our noise PEF estimation to be impacted by any signal in the estimation model. A better choice would be to use horizontal geophone data, but the data collected to complement this record is of insufficient quality to be used as a model for PEF estimation. Though not shown here, results using geophone data as a model for noise PEF estimation contain considerably more coseismic noise. The final result (Figure [*]b) shows that the signal noise separation effectively removed the bulk of the coseismic energy, leaving two interface response events in the first 0.01 seconds of the record, and predominantly random noise in the lower part of the record.


next up previous print clean
Next: Conclusions and future work Up: Haines and Guitton: Electroseismics Previous: Synthetic Electroseismic data
Stanford Exploration Project
7/8/2003