Providing Spatial Coherence Information to Fault Meta-Attribute

Summary Seismic fault interpretation is a very time-consuming manual task. Many attributes have been proposed as fault identifiers, most of which are based, directly or indirectly, on the data’s local orientation. Nevertheless, not one of these attributes alone is able to solve the automatic fault detection problem. They evaluate the degree of discontinuity along coherent events, but they can generate false positives in geological structures that are not faults. Furthermore, one aspect generally not considered by those attributes is the fault’s geometry, which typically corresponds to sub-vertical planes with low curvature. A robust approach to improve the quality of the detection is the so-called meta-attribute. It consists in a non-linear combination of a number of attributes with the use of a supervised neural network. In this work we propose a new attribute with low computation cost that seeks to identify the regions of the data that present a spatial coherence compatible with the fault’s geometry. The effectiveness of this proposal is demonstrated by comparing the classification performance of a network trained with just one traditional attribute and that of a network that also includes the new attribute in its training.