Usage of pre-stack vibroseis data before sweep correlation for noise removal and near-surface structural analysis

<p>We develop signal processing and imaging techniques for characterizing subsurface structure with utilizing raw vibroseis shot data before sweep correction. Vibroseis shots are often used for a land seismic survey, and the raw data of observation are typically converted to impulse-shot records with correlation of shot sweep and stacking. However, as we demonstrated here, the raw data are useful for providing additional valuable information. One opportunity we have is to clean up the vibroseis shot gathers by removing traffic noise in the raw data. Traffic noise is one of the strongest kinds of noise in the frequency range overlapped to the vibroseismic shots, and this noise contaminates the shot gathers. Diversity stacking helps to improve the signal-to-noise ratio （SNR）, but we develop a more advanced, and traffic-noise-oriented, filter with a machine-learning approach. This filter improves the SNR of shot gathers. Another opportunity is for near-surface velocity structure modeling. Near surface is important to know for topography correction of imaging （static collection） and shot and/or receiver couplings. However, the conventional vibroseismic shot geometry is not suitable to estimate these structures. We use ambient noise data recorded during the vibroseismic shot survey. Each sweep shot is 38 seconds and more than 10 sweep shots are repeated at each shot location. This means that we have 380 seconds of continuous data at each location. We use these data to extract surface waves using seismic interferometry, and estimate near-surface S-wave velocities from the surface waves. The velocity image provides detailed information of the structure at top 50 m along the entire survey.</p>