Earthquake Occurrence and Reoccurrence at Intermediate Depth in the Bucaramanga Nest, Colombia

Abstract:

We used a hybrid deep learning – template matching approach to create a catalog of more than 150.000 earthquakes originating in the compact Bucaramanga Nest of intermediate-depth seismicity in northern Colombia. We re-visited known earthquakes from the catalog published by the Colombian Geological Survey Seismological Network (Servicio Geológico Colombiano, SGC - Red Sismológica Nacional de Colombia RSNC) and used deep-learning-based phase picking methods, PhaseNet and EQ Transformer, to develop a consistent set of arrival times. We used the retrieved P and S arrival times to create a comprehensive set of templates for a subsequent template matching search. We apply this workflow to continuous data spanning the period January, 2012 to October, 2019 on the four closest SGC stations to increase the earthquake catalog from 50,000 events to more than 150.000 events. We use waveform similarity and precise event location to group earthquakes into repeating or near-repeating sequences, and highlight the occurrence of opposite polarity repeating earthquakes. The behaviors of these sequences is highly variable. We identify some sequences that persist from weeks to months to years, and with inter-event times that range from minutes to months. Some of the events include opposite-polarity events. We search the improved catalog to explore the recurrence intervals, relative moments, focal mechanisms, and locations of these earthquake sequences, to help constrain the mechanics of Bucaramanga Nest seismicity.