Simultaneous vibroseis data separation through sparse inversion

Abstract

Compared with conventional seismic acquisition methods, simultaneous-source acquisition utilizes independent shooting that allows for source interference, which reduces the time and cost of acquisition. However, additional processing is required to separate the interfering sources. Here, we present an inversion-based deblending method, which distinguishes signal from blending noise based on coherency differences in 3D receiver gathers. We first transform the seismic data into the frequency-wavenumber-wavenumber domain and impose a sparse constraint to estimate the coherent signal. We then subtract the estimated signal from the original input to predict the interference noise. Driven by data residuals, the signal is updated iteratively with shrinking thresholds until the signal and noise fully separate. We test our presented method on two 3D field data sets to demonstrate how the method proficiently separates interfering vibroseis sources with high fidelity.