Thousands of Induced Earthquakes per Month in West Texas Detected Using EQCCT

Abstract

West Texas has been a seismically active region in the past decade due to the injection of industrial wastewater and hydrocarbon exploitation. The newly founded Texas seismological network has provided a catalog that characterizes the intense seismicity down to a magnitude of 1.5 Ml. However, there are numerous small-magnitude events (Ml < 1.0) occurring every day that are not analyzed and reported, due to the prohibitively high workload to manually verify the picks from automatic picking methods. We propose to apply an advanced deep learning method, the earthquake compact convolutional transformer (EQCCT), to unleash our power in analyzing hundreds of small earthquakes per day in West Texas. The EQCCT method is embedded in an integrated-detection-and-location framework to output a highly complete earthquake catalog, given a list of available seismic stations, in a seamless way. The EQCCT has enabled us to detect and locate 50-times more earthquakes (mostly smaller than magnitude 1) than we previously could. We applied the EQCCT-embedded detection and location workflow to the Culberson and Mentone earthquake zone (CMEZ) in West Texas and detected thousands of earthquakes per month for consecutively three months. Further relocation of the new catalog revealed an unprecedentedly high-resolution and precise depiction of shallow and deep basement-rooted faults. The highly complete catalog also offers significant insights into the seismo-tectonic status of the CMEZ. Association with nearby injection activities also revealed a strong correlation between the rate of injected fluid volume and the number of small earthquakes.

Keywords

• deep learning
• seismology
• phase picking
• induced seismicity
• earthquake monitoring