A High-Resolution Decadal Earthquake Catalog for Oklahoma: Development and Interpretation

Abstract:

The non-stationary character of induced seismicity and the temporally varying seismic networks in Oklahoma impact the performance of the template matching approach for building a high-resolution catalog of earthquake hypocenters. A promising alternative for the task is to pick phase arrival times using a trained neural network and to use standard phase association, location, and relocation routines. This way, the usable data and the detectable earthquakes are not limited by the templates, which gives us the potential for finding new fault structures and finding more earthquakes. Like other phase measurement techniques, neural networks can suffer from missed (false negative) or misidentified (false positive) results, which complicates the phase association step. Here, we complement these limitations by checking for the double-difference relocatable conditions – correlated waveforms with similar moveouts – in a pairwise manner while associating the phases. This additional layer of strong constraint allows us to associate the events that were captured by a small subset of the stations, hence complementing the issue of false negatives, while screening out false positives. Using this approach, we develop and report on a high-resolution catalog for the past decade in Oklahoma and explore the insights that the new catalog can provide – from the possible role of fluids in earthquake occurrence to the geometric complexity of activated faults and their orientation with respect to the ambient stress field.