Abstract
SUMMARY
Distributed acoustic sensing (DAS) acquisition is becoming more and more popular for its dense sampling at a lower cost than seismometers. However, data processing for DAS data is challenging, especially for surface-deployedfibers, in which only the horizontal component of strain variation is effectively recorded. Also, the coupling between the fiber and the Earth is usually poor and the recorded single-component data are noisy. Thus, we introduce data processing strategies dedicated to enhancing the ambient-noise and active-source seismic data recorded by a horizontally deployed tactical fiber-optics cable buried in a sand dune area in Saudi Arabia. We propose a similarity-weighted stacking of randomly selected short-time duration windows to generate virtual common shot gathers (CSG) from the recorded ambient noise. The similarity-weighted stacking only counts the primary contributions of coherent events, while a short-time correlation can suppress the crosstalk usually present in late arrivals. The stacking fold is preserved or even can be increased by generating plenty of random time segments compared to stacking the full recording time. For the recorded active-source data, we skip the interferometric step, but use the envelope of the CSG. The envelope is needed to mitigate the complexity of waveforms, while preserving the slopes of arrivals. Then, we use the wave-equation-based Rayleigh-wave dispersion spectrum inversion, which utilizes all the dispersion modes available and does not require picking the dispersion curve, in estimating the shallow S-wave velocities. The local cross-correlation objective function allows for additional freedom in matching the modelled and observed data, and thus, helps us avoid falling into a local minimum when starting with kinematically poor velocity models.