A deep-learning based seismicity catalog for Northern Chile(2007–2023) containing >2 million events

Northern Chile, located at the boundary of the South American and Nazca plates, is one of the most seismically active regions in the world. To better understand the seismicity and tectonic processes of this complex subduction zone, we analyzed and processed data from the Integrated Plate Boundary Observatory Chile (IPOC) as well as all available temporary seismic stations, spanning the period from 2007 to 2023 (total: 243 seismic stations).

Using EQTransformer, a deep learning-based phase picker, we identified P and S wave arrivals with high precision across a vast dataset of seismic waveforms. We originally utilized the version of EQTransformer pre-trained on the INSTANCE dataset (available from SeisBench), but achieved better results by applying transfer learning based on hand-picked IPOC data. In total, 93,721,745 P- and 71,296,129 S- phases were obtained in this step. The selected phases were then processed with PyOcto, an advanced association and location tool, to group the phases into seismic events. This workflow resulted in a catalog of ~2.5 million events, about 10 times as many as the most complete regional catalog to date. Finally, we relocated the catalog using first Simul2000 and a 2D velocity model, then hypoDD to obtain relative locations, which provide a detailed view of the seismicity in the region.

We present a summary of the retrieved catalog, as well as zooms into potentially interesting subregions. Our catalog offers the potential for numerous follow-up studies, e.g. in statistical seismology or seismic tomography.