| dc.description.abstract | Seismogenic fault geometries are important to understanding the seismic potential of fault slip in natural systems. Their orientation relative to the principal tectonic stress directions control their susceptibility to slip. While fault orientations can be ascertained through detailed field mapping or subsurface imaging, rapid automatic identification of fault structures may be useful for future seismic hazard analysis. We apply a methodology workflow that can be automated and utilize central Oklahoma seismicity since 2010 as a test of the concept. To conduct this retrospective analysis, we utilize a Python package, easyQuake, that was recently developed by the Oklahoma Geological Survey (OGS), to augment their earthquake monitoring via pickers that are trained on millions of seismograms with neural network deep learning. The easyQuake software was used to analyze seismic data collected from 2009 to 2015 - specifically to investigate the possible migration of earthquake nucleation in the Jones, Oklahoma region - as well as increase the catalog completeness and accuracy. Jones is the area of interest for this study as it was an early case study for large volume (~1 Mbbl/mo) wastewater injection inducing earthquakes at greater than 10 km distances. The software outputs a catalog in QuakeML format, of which the associated metadata, such as pick times, can be read into hypoDD format. The catalog is then relocated with hypoDD to explore cluster and fault geometries, as well as earthquake sequence migration. With the hypoDD results, we utilize machine-learning techniques packaged in the Python machine-learning library scikit-learn for spatial clustering (e.g., DBSCAN) followed by linear regression (e.g., RANSAC) of those clusters to identify suspected fault segments. We identified potential instances of previously unmapped, seismically active segments in the Jones area and plan to expand the analysis to examine a wider range of the state’s fault segments and more current seismic sequences. Many of the segments mapped would not have been identifiable without the expansion of the microseismic catalog easyQuake provides. This type of work improves our understanding of the seismic hazard as it increases the database of known fault segments throughout the broader region and shows that these seismogenic fault segments can be identified automatically. | en_US |
