Predictive Coherency

Karimi, Parvaneh (The University of Texas at Austin) | Fomel, Sergey (The University of Texas at Austin)



Detection and interpretation of fault systems are crucial to seismic interpretation and reservoir characterization. We introduce a new attribute that aims at detecting faults while preserving fault information and handling its local variations. First we use predictive painting to form a structural prediction of seismic events from neighboring traces. Then we compute prediction residuals and find the smallest prediction-error interval at each point that is the best representative of fault information at that point. In comparison with other fault attributes, such as classic coherency or similarity, predictive coherency allows for a balance between highlighting of faults and protection of fault information and its local variations. To asses performance of the proposed attribute in highlighting faults, we compared results from our attribute with analogous attributes over the same dataset. The comparison demonstrates the effectiveness of fault detection using predictive coherency.