Subsurface Insights March 2021 | Page 15

Subsurface Insights | 15
TRADITIONAL METHODS FOR IMAGING FAULTS IN SEISMIC
To identify faults in seismic , interpreters typically scan laterally across a vertical seismic section , looking for either high discontinuities or low continuities along seismic reflectors . In the years since the first seismic interpretation software was introduced , many seismic attributes ( additional volumes calculated from the input amplitude volume ) have been proposed that algorithmically highlight these anomalies and generate fault-image volumes that interpreters use to assist in their interpretation workflow . Approaches based on discontinuity include Bahorich and Farmer ( 1995 ), while approaches based on continuity ( e . g . semblance ) include Marfurt et al . ( 1998 ). Examples of these attributes can be seen in figures 1B and 1C .
While these attributes are an improvement over using original amplitude volumes for interpretation , they fall short of the quality needed to automate fault plane extraction . This is due to their noisy nature and characteristic
of picking up other non-fault features , such as stratigraphic changes or unconformities . Many of these issues are clearly illustrated in figures 1B and 1C . While more sophisticated approaches based on these methods have been developed , they still exhibit many of the same limitations .
The introduction of the fault likelihood attribute ( Hale , 2013 ) improved fault interpretation significantly . In this approach , semblance is calculated within an elongated , “ fault-like ” window , which is rotated around multiple strike and dip orientations to identify the minimum semblance ( the maximum likelihood ) for a fault ( Figure 2 ). This process generates a fault image volume of significantly higher quality than the traditional fault attributes , but can still appear blurry and less geologic than desired ( Figure 1D ). Additionally , a “ thinning ” step looks at the fault likelihood images and defines a ridge that corresponds to the most likely location of a fault ( Figure 1E ).
As we can see from figures 1D and 1E , the fault likelihood attribute gives a much clearer image
Figure 2 > Simplified illustration of fault likelihood calculation . For each sample within the seismic , a semblance calculation is performed within an elongated , fault-like window . That window and calculation is rotated around multiple strike and dip orientations to find the lowest semblance ( the highest fault likelihood ). This is repeated for every sample in the dataset .