4D Deblending Using Sparse Inversion

Abstract Simultaneous-source seismic acquisition improves acquisition efficiency. However, strong blending interference poses challenges for deep and ultra-deep seismic exploration, where weak-signal preservation is critical. Most industrial sparse-inversion-based deblending methods operate in the three-dimensional (3D) common-receiver gather (CRG) domain. These methods process each CRG independently, without making use of signal coherence along the receiver-inline direction. Exploiting additional signal coherence in higher dimensions has the potential to improve seismic signal processing performance and weak-signal preservation. In this study, we develop a four-dimensional (4D) sparse-inversion-based deblending framework that exploits receiver-inline signal coherence. Here, 𠇄D” refers to a processing volume with dimensions of receiver-inline, source-inline, source-crossline, and time, rather than conventional time-lapse seismic. The proposed method jointly processes multiple neighboring CRGs from the same receiver line and employs a windowed 4D fast Fourier transform (FFT) to obtain a higher-dimensional sparse representation of the seismic wavefields. By incorporating receiver-inline coherence into the inversion, the 4D method provides a sparser representation of coherent seismic signals and helps separate weak signals from strong blending interference. Synthetic data and blended ocean-bottom-node (OBN) field data examples show that the proposed 4D method achieves better deblending performance compared with the 3D sparse-inversion-based method. These results indicate that using high-dimensional signal coherence can provide a simple and practical improvement to existing deblending workflows.