Tracking Induced Seismicity in the Fort Worth Basin, Texas and Northern Oklahoma Using Local and Large-N Style Arrays

Since 2008, earthquake sequences within the Fort Worth Basin (FWB), north Texas, have been linked to wastewater disposal activities related to unconventional shale gas production. Here, I present my work analyzing and cataloging a complete record of the seismicity occurring within the basin over a period of more than a decade (2008-2020). Analysis of the catalog reveals that the earthquakes generally occur within the Precambrian basement along steeply dipping normal faults, and while overall seismicity rates have decreased since 2016, new faults have become active. I observe strong spatial and temporal correlations between the earthquake locations and wastewater disposal well locations and injection volumes, implying that fluid injection activities may be the main driving force of seismicity in the basin. To analyze the basin-wide injection associated stress changes, I created a coupled geomechanical model of the Fort Worth Basin incorporating geology and fault data shared by our collaborators at TexNet and the Bureau of Economic Geology. Models exploring endmember geologic and tectonic scenarios were created using the COMSOL modeling software and are fully coupled, meaning that both injection associated pore fluid pressure and poroelastic stress changes are calculated. The preferred model shows that while large injection associated stress changes are concentrated in those portions of the basin where both injection volumes and rates are highest, there are observable far-field stress changes capable of reaching faults at far distances (\u3e15 km) away from injection wells. Further work focusing on the Dallas-Irving sequence, located \u3e15 km from low-to-medium volume injection wells, involved the creation of a high-resolution relative location catalog and template matching work to respectively better estimate the spatiotemporal characteristics and start date of the sequence. The results indicate that the Dallas-Irving sequence likely began in very late 2013 or early 2014 and that the seismicity associated with the sequence is occurring along a NE-SW trending normal fault. Importantly, the Dallas-Irving sequence only became active following the arrival of injection associated stress changes at the sequence site. Lastly, the dissertation shifts focus to explore newer earthquake detection techniques to understand their usefulness in monitoring and creating earthquake catalogs efficiently and rapidly. The work focuses on the Community Wavefields Demonstration Experiment (CWDE) data set in northern Oklahoma. The local similarity event detection method utilizes the large-N style nodal array to detect events without the need for a priori information related to the geology of the study region. The study focuses on the creation of a benchmark earthquake catalog made using traditional analyst-based analysis of the broadband seismic data collected as part of the community experiment. This benchmark catalog is used for comparison with the local similarity nodal detection list. Ultimately, the local similarity technique was able to detect nearly all (97.0%) events recorded within the benchmark broadband catalog and was also able to detect over 1000 events which were not included in the benchmark catalog. Through the analysis of the local similarity results I was able to extract network-wide first P- and S-wave arrival onset time estimates which were used to discern local and regional distance events from the array. The NTXES catalog is the most complete record of seismicity occurring in the FWB and represents a great starting point for possible future analysis of long-spanning induced seismicity. Continual monitoring of seismicity in the FWB yields insight into both the lifespan of induced seismicity sequences, and also tell us about how seismic sequences may end. Additionally, the evolution of the FWB seismicity could inform us about the effectiveness of injection mitigation techniques considering that, due to a variety of factors, injection activities have gradually been decreasing since their peak in 2012, and likewise seismicity rates have continuously decreased following their peak in 2015. For monitoring purposes, the created event detection list of the CWDE data set serves as a great starting point, but further analysis steps need to be undertaken to convert these results into a full earthquake catalog. Further analysis of the CWDE data set may be extended to other event detection and phase picking techniques and algorithms which could either seek to take advantage of the benchmark catalog for approaches such as template matching or machine learning or instead focus entirely on the nodal array data and benchmark their results on the local similarity results instead. Ultimately, the CWDE data set remains an excellent data set for the purposes of testing a variety of event detection and phase picking techniques which could be incorporated into the workflow analysis of any future network deployment involving the installation of large-N style broadband and nodal station arrays