Deblending of Flip-Flop OBN Data in Shallow Water

Separating energy from different seismic sources, or deblending, is commonly applied in land and multi-vessel marine surveys. A shallow-water dense OBN (Ocean-Bottom Node) dataset acquired in a standard flip-flop mode is a challenging candidate for deblending as the firing time-distance rule there is optimized for higher production rate rather than for deblending-friendly dithering. A special deblending approach improving the primary signal quality and relying heavily on structure of the data has been developed and applied here. The implemented source separation technique is iterative and relies on some additional information on the signal properties (priors). At each stage, the algorithm makes a prior-guided effort to isolate various components of the seismic signal, including direct arrivals, reflected waveforms, refracted signals, and diffractions. These elements require identification and windowing for accurate separation in the sparsity domain. The process starts with the most prominent signal component, usually the direct arrival or Scholte wave, and then iteratively applies tailored exit strategies to maximize signal preservation and fidelity while maintaining the integrity of the original waveform. The resulting deblended common receiver gathers show an excellent recovery of the first arrivals in full frequency band, including low frequencies. Continuous and undisturbed first arrivals are readily available now for the Full Waveform Inversion (FWI) velocity model building step, which is a very important achievement of this study. The steeply dipping linear energy associated with the slowly propagating Scholte waves from the earlier shots was successfully removed by the algorithm. A strong noise associated with the following shots was almost totally suppressed. The resulting stacked image shows a nicely recovered primary reflectivity that became visible after the separation making interpretation much easier. The in-field dithering randomization has been increased following the recommendation of this study. The novelty of this study is the ability to recover primary source energy in a flip-flop single vessel OBN dataset acquired with limited dither and high shot density. This was achieved by careful windowing of different modalities of the signal and applying specific thresholding criteria.