Removal of Trigger Delays from Cross-borehole Seismic Data by Exploiting Tube Wave Coherency – A Pre-processing Tool for Waveform Inversion

Abstract:

A cross-borehole seismic dataset was acquired using two 50 m deep test boreholes drilled into Permian age sandstones near Mels, Switzerland. The boreholes were approximately 20 m apart, and useful data frequencies between 100 Hz and 4000 Hz. The survey comprised 181 source locations and 192 receiver locations, using multiple deployments of a 24-element receiver array to ensure adequate spatial sampling. Imaging velocities and attenuation using Full Waveform Inversion (FWI) proved difficult due to the presence of significant trigger delays in the survey.

Trigger delays can be solved for using either a Traveltime Tomography (TT) approach (by including additional unknowns), or a FWI approach (by individual source estimation). However, these methods introduce significant ambiguities, particularly in the presence of an unknown level of anisotropy. Moreover, the trigger delays break coherency in the common source and common receiver domains, preventing simple removal of the tube wave energy in the data.

We introduce a novel method for estimating the trigger delays through cross-correlation of tube waves. Repeat records from identical source-receiver array combinations are shifted and then stacked to create a “reference dataset”. We make use of the high amplitude and highly coherent tube waves, which travel at a constant velocity: receiver hole tube wave arrivals are aligned and enhanced by applying a linear moveout and an f-k filter to the common source gathers. Each physical source location is recorded by four closely spaced and interleaved receiver elements; the high level of redundancy enables recovery of a stable set of delay estimates. A similar approach is used for the common receiver gathers to remove the remaining delays between source locations.

Trigger delays were corrected with a mean absolute value of 153 μs, a mean bulk shift of −41 μs and a standard deviation of 75 μs. Their removal from the raw seismic data, allowed i) the tube waves to be removed by standard f-k processing, ii) the resulting (shifted) waveform arrival picks to be used for anisotropic TT, and iii) FWI to be successfully applied to the pre-processed data.