Abstract
Shooting in an overlapping fashion is called blended or simultaneous acquisition. This can make acquisition efficient and economical. However, deblending (i.e. the process of separating the blended data) of data can be challenging.
Blended seismic data can be acquired in a number of different configurations. Examples are a shooting vessel follows the tail buoy of a conventional streamer vessel or a shooting vessel and conventional vessel are parallel. One can also vary the signal strengths and frequency bands of the sources for the two vessels. In this way both shallow and deep resolution can be improved as low frequency waves penetrate deep as compared to high frequency waves.
In this thesis I have tried various conventional seismic processing techniques (like tau-p transformation, FK filtering, SVDMUL, Time-frequency Denoising, and FX prediction) to deblend or separate the blended data. I used several datasets (one real and 2 synthetic), with different geometries, but all the datasets have same acquisition design in which one vessel is in front of the streamer and the other is behind the streamer cable.
I tried to separate or deblend the data either on the bases of their dips, or alternatively by transforming the deblending task into that of denoising. Events related to the shooting vessel appear as random noise in Common depth (CDP), common receiver (CR) and Common offset domain (CO). SVDMUL and TFDN methods are then applied in the CO and CDP domains respectively.
I found that the best deblending results were obtained using SVDMUL (Singular-value decomposition approach) for all data sets. SVDMUL is used iteratively and the deblended result is gradually built up. This method performed superior compared with deblending based on TFDN, FK or tau-p coherency filtering. The tau-p transformation based method showed significant problems in areas of strong or conflicting dips. The efficiency of TFDN and SVDMUL methods are based on the time jitter between the shots from to the conventional vessel and shooting vessel. If too small the techniques from worse.