Characterization of Microseismicity of the Rogersville Shale
As part of the Conasauga Shale Research Consortium sponsored by the U.S. Department of Energy, John Hickman, Carpenter, Wang, and Schmidt operated seven sensitive seismic stations in the Rome Trough of eastern Kentucky beginning in October 2019. These stations were installed in 2015 as part of a pilot study to characterize microseismicity in areas where the possibility of inducing earthquakes by producing hydrocarbons from the Rogersville Shale is greatest and where clusters of wastewater injection wells operate. A new seismic station was installed near Paintsville in February 2020. Recordings from this station and a network of stations are acquired in real time and analyzed at KGS as a continuation of the investigation that began in 2015.
Carpenter, Wang, Woolery, and Hickman, along with Andrew Holcomb (KBRWyle Technology Solutions Inc.) and Steven Roche (University of Tulsa) published results from the initial microseismicity study, “Natural Seismicity in and Around the Rome Trough, Eastern Kentucky, From a Temporary Seismic Network,” in Seismological Research Letters in May 2020. The study found that the Rome Trough of eastern Kentucky appears to be seismically quiet: only three earthquakes were observed during the study period in the crust bounded by the trough’s boundary faults, where the temporary network was most sensitive. This finding contrasts with findings for areas of much greater activity north and south of the trough, including the source regions of the 1980 magnitude-5.0 Sharpsburg and 2012 magnitude-4.2 Perry County earthquakes, and suggests that the earthquake potential in the Rome Trough is low.
Machine-learning techniques were applied to analyze seismograms from the initial microseismicity study to accurately and automatically distinguish between earthquake recordings and mine blasts. More mine blasts than earthquakes were recorded by the seismic network in eastern Kentucky and Tennessee, southeastern Ohio, southwestern Virginia and West Virginia, and western North Carolina between 2015 and 2018. The results were published in “High-Accuracy Discrimination of Blasts and Earthquakes Using Neural Networks with Multiwindow Spectral Data,” by Carpenter, Wang, Holcomb, Woolery, and visiting scholar Fajun Miao in May 2020 in Seismological Research Letters. Machine-learning techniques can greatly improve the efficiency of data analysis and reduce an analyst’s effort for event classification by as much as 97 percent.