LAUSR

Abstract Aftershocks of the 16 September 2015 M8.3 Illapel earthquake in central Chile were recorded for a period of one year by the Chile-Illapel Aftershock Experiment (CHILLAX) seismic network. An initial catalog of about 100,000 events was generated by a novel automated picking algorithm that combines an auto-regressive detection/onset estimation method with some recently developed windowing techniques. We combine arrival times of P and S waves from about 9,000 of the best recorded of these events with relative arrival times of teleseismic P waves and phase delays of Rayleigh waves recovered from ambient noise to generate a three-dimensional image of P and S wavespeeds in that part of the Andean margin located beneath the network. Hypocenters of aftershocks located in the final model appear to be grouped into four distinct zones: two lower parallel zones that dip about 20° to the east, an upper, highly active zone that dips about 30° to the east, and a diffuse zone of smaller events located between this upper zone and the surface. Nearly all of the events with M > 4 are located at the bottom edge of the highly active zone and define a plane dipping about 26° to the east. Most of these events have almost identical thrust mechanisms and are separated from the lower zones by a steep gradient in increasing wavespeeds. The lower end of the upper zone, at about 55 km depth, is marked by another velocity gradient dipping to the west that bounds the diffuse shallow activity and projects to the surface west of the high Andes. We postulate that the triangular region delimited by these two high wavespeed gradients is a subduction wedge, generated by the removal of material from the leading edge of the upper plate by subduction erosion, and subsequently underplated back on to the overriding plate at depth. This model explains many of the features we observe in the seismicity and wavespeeds and is consistent with geologic observations of uplift and extension in this part of the Chilean coastal margin. Characteristics of this wedge, such as its shape and the lack of correlation with the coeval back-arc tectonic setting, suggest that high Andes are not directly coupled with the behavior of the subduction interface.