Survey of algorithms for automatic detection of fractures from BIG spatial Datasets: One of the important aspects of the research to be pursued in the project is the delineation of characteristics of natural fractures from very large 3D seismic data sets. While traditional seismic processing algorithms such as Residual Move-Out Analysis and pre-stack seismic inversion can yield valuable information regarding fracture strike and variations in shear and compressional velocity in terms of azimuthal variations, relating spatial variations in these quantities to actual fracture characteristics will require novel algorithms for pattern recognition and pattern simulation constrained to other physical constraints. Research in this area has been initiated by a graduate student funded through the NSF BIG Data project. The plan is to engage an undergraduate researcher to review and implement other traditional algorithms for feature identification that have been proposed in the literature. That student will implement more traditional workflows such as coherency analysis and semblance analysis that have been employed in order to delineate features such as faults and major fractures in seismic data sets. The implementation of some of these algorithms will also serve to benchmark some of the methods that will be developed as part of the BIG DATA project.

Survey of algorithms for pattern recognition from BIG time series data: Fractures control the migration of fluids in the subsurface. Sudden changes in fluid migration pathways due to the presence of fractures in turn manifest themselves in the form of subtle variations in the injection rate profile during injection or in the form of variations in pressure recorded at monitoring wells. Understanding and quantifying these time-series variations is therefore crucial for gaining an insight into the characteristics of fractures in the vicinity of injection locations and within the subsurface geologic formation in general. A systematic assessment of the production/injection characteristics at wells corresponding to variations in fracture characteristics such the spacing, clustering intensity, aperture variations is being conducted by a graduate student working on the BIG Data project. The plan is to engage a second undergraduate researcher who will investigate and apply more traditional techniques for automatic detection of anomalies in BIG time series datasets. These could include wavelet analysis, Fourier analysis and analysis of the power law characteristics of temporal signals. This investigation will nicely compliment the efforts to develop fast data assimilation schemes that will yield realistic models for fractures in subsurface geologic formations.