Seismic Characterization of Faults from 3-D Seismic Data: Fault Detection, and Characterization of the Bakken Formation

This dissertation focuses on automatic fault detection and generation of different fault attributes from a seismic discontinuity volume such as dip and strike, density, displacement, and rose diagrams to characterize faults from 3D seismic data in the Bakken Formation. I use principal component analysis (PCA) for automatic fault detection. The PCA fault attribute indicates significantly different, and geologically more plausible, 3D fault distributions than do conventional seismic attributes, such as curvature. The PCA fault attribute identifies different fault patterns in the Upper, Middle, and Lower Bakken members and the Three Forks Formation. Two distinct fault trends in approximately N40–50°E and N50–60°W directions are observed in the Bakken Formation. Fault cuts interpreted from missing well-log sections also correlate well with the PCA fault attribute. Seismically derived fault orientations correlate with borehole image logs in horizontal wells. Crossing conjugate faults observed on the fault-dip attribute are interpreted to result in widening of the faulted area and localized thinning of the rock sequence at the intersection of conjugate faults. This fault network potentially enhances permeability in the direction of fault strike. I demonstrate the potential for using the time-frequency continuous wavelet transform (TWCFT) for detecting small faults in the thin Bakken Formation. TWCFT, along with PCA, detected faults with different vertical displacements which are preferentially highlighted at different dominant frequencies. At the Bakken level, major faults are seen best at low frequencies, while antithetic faults around the master faults are seen best at high frequencies. I analyze rose diagrams from seismic to identify different patterns of preexisting faults in ~N45°E, ~N75°E, and ~N15°E orientations for normal and strike-slip faulting stress state. The rose diagram results correlate well with rose plot generated from image logs. Fault analysis of vertical sections illustrates fault connectivity from the Bakken to the adjacent formations. Finally, density-porosity predicted with multi-attribute analysis of seismic data is correlated to fault density.