Automatic detection of seismic phases using variance fractal-dimension trajectory

Abstract

Seismic refraction signals in crustal surveys are often contaminated by various types of noise mainly due to low-signal/noise Earth environments. A variance fractal-dimension technique is investigated and applied to real data sets for detection of seismic refraction signals from background noise. The sharpness of transition features on the variance fractal-dimension trajectory was used as a criterion for distinguishing the seismic signals, and a window size of 48 samples and a window spacing of 8 sample intervals were chosen to calculate the fractal dimensions and to create the trajectories for subsequent tests and detection of phases Pg, Pn, PmP, and ground roll. The real data tested in this study were obtained from the 1992 Abitibi—Grenville Lithoprobe high-resolution refraction and wide-angle reflection experiments. The transition criterion provides a robust and powerful tool for separating the real seismic-refraction signals from the background noise. In the case of Pg phase from the refraction profile fan-shot ab2, the variance fractal-dimensions are generally smaller for the refraction signals (≈1.25) compared to the background noise (≈1.75).