Full text loading...
-
Low-frequency Compensation and Long-wavelength Velocity Inversion Using Sparse Bayesian Learning
- Publisher: European Association of Geoscientists & Engineers
- Source: Conference Proceedings, 79th EAGE Conference and Exhibition 2017, Jun 2017, Volume 2017, p.1 - 5
Abstract
Low frequencies have been demonstrated to be very useful in seismic processing, inversion and interpretation. However, seismic datasets acquired via early seismic survey is devoid of low frequencies (e.g., 0–8 Hz). Although the frequency bandwidth can be extended down to ∼1.5 Hz via modern expensive acquisition technique, the existence of multiple wave, surface wave and random noise in seismic data readily damages the available weak low-frequency signals. Therefore, developing processing methods to boost low-frequency signals is often necessary as well. The lack of low frequencies results in some false weak signals, and thereby causes the subsurface sparse reflectivity series non-sparse. The sparse Bayesian learning (SBL) method can invert for a sparse reflectivity series from bandlimited data, due to the utilization of a hierarchical Gaussian priori. Consequently, SBL reflectivity inversion has the ability of retrieving low frequencies. The long wavelength components of velocity, which is usually a good initial model for generating the absolute velocity, can be subsequently extracted from the recovering low-frequency information by using the conventional velocity inversion. Synthetic and real borehole-side seismic data examples illustrate that the method combined SBL reflectivity inversion with velocity inversion can provide credible long wavelength components of velocity.