Investigating the regional and local structure of Oklahoma's crust using induced earthquakes

Abstract

There are still fundamental questions that remain unanswered with respect to the processes involved in the formation and evolution of the continental crust of North America. Data about the rock properties and rock types from the deeper crust can help us answer these questions. Seismic data provides one of the most detailed looks at the characteristics of the rocks in the subsurface. Active tectonic margins are well studied as these regions have a lot of natural seismicity due to active tectonic processes. On the other hand intracratonic regions can be difficult to study due to the lack of such natural seismicity. Deep crustal active seismic studies can be expensive and spatially restrictive. Oklahoma presents a unique case as the intraplate seismicity due to waste water injection provided the data required to evaluate the deep and shallow crustal structures. The combination of high local seismicity, installation of local monitoring stations in response to it, and coincidental overlap of the US Transportable Array can be utilized to study the deep crust of Oklahoma.

To study the deeper crust, I present a non-standard methodology of processing a passive seismic data set. Vertical and horizontal component data with a maximum offset of 250 km is selected for processing. The data are bandpass filtered, converted to envelope, and STA/LTA (short-term average/long-term average) ratio applied. Next, the waveforms are sorted into common-mid-point (CMP) gather and stacked into 5 km offset bins. These steps simplify the waveforms and increase the signal-to-noise ratio, thus making it easier to identify and pick the P- and S-wave arrivals, especially for far-offset(>150 km) traces. Travel-time curves for each of the CMP bin are picked and inverted to obtain 1-D velocity-depth function at respective CMP bin locations. These are then combined to obtain 3-D P- and S-wave velocity model for the crust of Oklahoma up to 40 km depth. The P- and S-wave velocity models are combined to produce a Vp/Vs ratio and Poisson's ratio model for the crust. The results are enlightening and show a high P-wave velocity (>7 km/s) lower crust for Oklahoma. Vp/Vs ratios of >1.8 are reported for the lower crust. These values suggest a mafic lower crust that possibly formed through mafic underplating and crustal melting. The P- and S-wave velocity anomalies observed in the upper-middle crust are well correlated with local structures and gravity data. This methodology overcomes the limitations of traditional local earthquake tomography where the low S/N ratio for far offsets limits the depth of investigation. Our final velocity models reveal a heterogeneous upper crust that transitions to a more homogenous lower crust.

The source of induced seismicity in Oklahoma has been the waste-water injection operations related to the local petroleum production activities. To understand the earthquake source processes and fault dynamics we need accurate earthquake location estimates. The vast majority of earthquake location algorithms improve on the lateral earthquake location but still have major inaccuracies in the earthquake depth estimates. I use delay times observed between seismic phases to improve depth estimates for the local basement structure and the observed earthquakes. A local nodal array deployment over the Cushing Fault Zone in Oklahoma recorded three earthquake events that occurred on the fault that show seismic phase conversions possibly occurring at the basement interface. Using travel time modelling of the phase delay times I can establish that the observed phase conversion is an S-to-P (Sp) conversion at the basement and a basement depth of 1.05 km is estimated using the delay times between Sp and S phase arrivals. Using the established basement depth, I model synthetic waveforms for the three events for varying earthquake depths. Finally, I pick the P and Sp arrivals on modelled and observed data and use their delay times to estimate the earthquake depths. The results for basement depth and earthquake depth do agree with the local geology and seismicity. This technique can be used to improve the local structure and earthquake depth estimates. I also propose future work that involves full waveform modelling for each of the earthquakes that will remove assumptions related to the moment tensor of the earthquakes and will lead to a more accurate depth estimates.