Gaussian process regression in seismic fault detection

Gaussian process regression, as a nonparametric probabilistic model based on Bayesian statistics, is highly capable of supporting ‎sparse features such as global anomalies. Detecting abnormal behavior from normal behavior makes Gaussian process regression ‎as an edges detector where faults may occur in the seismic data. In this study, the Gaussian process regression-based anomaly ‎detection was applied to both synthetic and real data containing normal fault to detect the fault edge. To identify the fault edges, ‎the geological layers are considered as normal interaction and the fault edge as a global anomaly which disrupts the normal ‎behavior of layers. The error of regression is analyzed to separate the fault edge. To evaluate the proposed method, it was applied ‎on a series of synthetic seismic data and a real 2D seismic section of F3 block of the North Sea containing the fault. The results ‎show the ability of this method in fault detection.‎