Abstract. In this study we process 5 years of continuous seismic data, from 2017 to 2021, recorded by networks deployed along the Chilean convergent margin of western South America. We select a set of stations so that coverage is as continuous and consistent in time as possible. Thus, changes in the catalogue are mainly due to the processes responsible for the seismicity itself and not network artifacts. For phase picking, we use the deep-learning algorithm Earthquake Transformer, PyOcto for association, NonLinLoc for location and HypoDD for relocation. To handle the large spatial extent of the study area, we create a partitioning workflow that mitigates problems that arise from associator and location codes working in cartesian coordinates. This allows us to obtain a main seismicity catalogue with over 600 000 double-difference relocated events and a completeness magnitude of ∼2.5, with events located using an existing 3D velocity model (https://doi.org/10.5281/zenodo.15284375, Riedel-Hornig et al., 2025). While this provides better location accuracy than a 1D or 2D velocity model, it limits the final catalogue to the extent of the 3D velocity model which does not cover our whole study area. Therefore, we create a secondary seismicity catalogue for the region to the south of the 3D model. Our catalogues provide a dense, high-resolution dataset, with large spatial extent, for the interpretation of seismic processes along the western South American Margin.
Riedel-Hornig et al. (Mon,) studied this question.