Seismicity of the south-western South American margin through a machine learning automated approach

The South American margin, where the Nazca Plate subducts below the South American plate is a highly seismogenic region. In recent decades, it has been the focus of abundant studies with the deployment of temporal and permanent seismic networks. These efforts have generated large datasets that are challenging to process with traditional methods. To take advantage of the large volume of data available, we pair modern machine learning picking and association methods with traditional location and relocation techniques to create a dense, high resolution seismic catalogue. We process data from the CSN, OVDAS and other smaller permanent networks between 2017 and 2021 to obtain over 650.000 double-difference relocated events, at least 10 times more than any other regional catalogue in our study area. This implies at least a 1 order of magnitude reduction of the magnitude of completeness (Mc).

Our catalogue is designed to ensure temporal consistency (i.e. the selected stations are active for most or all of the study period) and the processing workflow is the same for the whole region (in contrast to joining catalogues resulting from independent local or regional networks). This consistency paired with the catalogue’s high resolution, allows us to observe spatial and temporal variations in seismicity and to improve our understanding of processes that may be studied through micro-seismicity. One application of such a catalogue is the observation that, although the Chilean subduction zone is known for its megathrust earthquakes, intraslab events make up the bulk of seismicity (>80% of the events), with two particularly active clusters. One is located in northern Chile, inland of the subduction of the Iquique Ridge, at an “unusual subduction segment” documented by Sippl et al. (2018) and another inland of the subduction of the Juan Fernandez Ridge. Furthermore, the slab and plate interface are most active in northern Chile an seismicity diminishes towards the south, especial from 36°S. This decrease in seismicity is likely related to changes in Nazca Plate age and temperature and/or to the influence of 1960 Valdivia and 2010 Maule mega-earthquakes.