APPLICATION OF AGDEBLEND IN SEISMIC DEBLENDING ON SEISMIC DATA LINE “X”

Abstract

The efficiency of seismic acquisition has encouraged the use of blended acquisition techniques, which produce overlapping data and require a signal separation process known as deblending. This study aims to evaluate how effective AGDeblend is in separating overlapping seismic signals in synthetically created blended data. The method used for deblending is an inversion-based approach, namely AGDeblend, and the effectiveness of the separation process is assessed. To achieve this goal, the methodology involves generating synthetic blended data from conventional seismic datasets (both marine and land). The signal separation process is carried out using the Fast Iterative Shrinkage-Thresholding Algorithm (FISTA). The performance of the method is evaluated quantitatively using Signal-to-Noise Ratio (SNR), Normalized Cross-Correlation, and Relative Error metrics. The main results show that the method performs very well on synthetic marine blended data, with an SNR value of 23.04 dB, a Cross-Correlation value of 0.99, and a Relative Error of 0.07. However, its performance decreases significantly on synthetic land blended data, where the SNR value is only 13.11 dB, the Cross-Correlation is 0.97, and the Relative Error increases to 0.22. This is characterized by signal leakage and less clean reconstruction results. The conclusion of this study is that the AGDeblend method successfully separates overlapping seismic signals. Furthermore, the effectiveness of AGDeblend strongly depends on the characteristics of the input data and how well the data meet the assumption of signal coherence. While it performs very well under ideal conditions, this study also highlights the challenges in applying the method to more complex data and emphasizes the importance of fulfilling the method's assumptions for successful inversion-based deblending.