Discussion

The processing sequence presented here is shown to be effective in separating the electroseismic interface response from the much stronger coseismic noise for the artificial example used. This is interesting both as an example of signal-noise separation in a case with very low signal-to-noise ratio and also as an important step toward proving the utility of the electroseismic method. However, the results we present are unrealistic in several important ways. Most importantly, the interface response events that we separate from the noise are essentially synthetic, and undoubtedly have different amplitude and frequency patterns than real arrivals from subsurface layers. We can attempt to make the inserted arrivals more realistic by stretching and scaling them before addition, but no synthetic record can really replace real data.

Thus the next logical step in this project is to collect a more realistic electroseismic data set at a location that is thoroughly-characterized. It is important that the characterization include wells, because the method promises to image layers that are invisible to existing surface methods. An ideal survey would take place at a test site with relatively simple subsurface, but with a few thin layers (fractures, clay lenses, etc) to be targeted. The use of a vibratory source or a high-intensity explosive source would vastly ease the recognition of the signal amidst background noise, and adding more recording channels would tremendously improve data quality and the output of data processing efforts.

A greater number of traces per shot gather will permit refinement of the processing scheme. The use of non-stationary PEF's may prove effective, since the nature of the coseismic energy clearly varies within the shot record. The use of a gap in the noise PEF may be an improvement, as it could prevent the PEF from being dominated by energy corresponding with infinite slope due to the spiky appearance of the coseismic data. In addition, a larger number of traces would improve the probability of success of transform-based methods such as f-k filtering and slant-stack filtering.

Our data processing goal is to develop a sequence that can be applied to field data in as close to an automatic manner as possible (fewer manual steps). This could include the development of generic signal and noise PEF's that would function on any data, and would not require the determination of these PEF's on windows of each data file. This would improve the processing of noisy data files (where such windows would be hard to define), and would speed the processing of data from larger-scale surveys. The results presented here suggest that this goal can be attained within a reasonable period of time, given a suitable data set for testing. The results of () demonstrate that such a data set can be collected in the proper setting with the necessary equipment. Thus, depending on circumstances, the electroseismic method could soon begin to provide useful new subsurface information in geophysical exploration.