High-frequency seismic radiation during Maule earthquake (Chile, 27/02/2010, Mw 8.8) inferred by backprojection of P waves: evidence of activation of two distinct zones at the downdip part of the plate interface

We back-project the seismic radiation released by Maule earthquake (Chile, 27/02/2010, Mw 8.8) in three frequency bands: 0.4-3 Hz, 1-4 Hz, 2-8 Hz. We measure the coherence of the seismic traces at 557 stations of US array by semblance. Travel times are estimated starting from a 1D global velocity model (ak135) corrected by two terms: a static correction and a dynamic correction. Static corrections are the mean time corrections to the 1D velocity model, and dynamic corrections are finer time shifts depending on the source-receiver path. Both terms are extracted from the time shifts between different receivers of P-phases of 23 high-magnitude calibration aftershocks, most of which have high precision locations based on the temporary deployment following the Maule earthquake (IMAD). The dynamic corrections are extended over a fine source grid by kriging interpolation. This procedure makes the backprojection results independent of the main shock catalog hypocentre and allows coherent imaging to higher frequencies. During the first 20 seconds of the rupture process, the source is stable nearby the nucleation point, which is close to epicentre proposed by Vigny et al (Science, 2011) based on high rate GPS motion. Afterwards, it moves bilaterally, with the northern front moving with an average velocity of ∼3 km/s. Most of the energy is emitted from the northern patch of the bi-lateral rupture (∼70%), with sporadic emissions from the southern patch. The maximum of stacked energy is located about 150 km north-eastwards from the epicenter and a relative maximum appears south of Arauco peninsula. In the dip direction, energy is mostly emitted from the down-dip edge of the co-seismic area, roughly matching the aftershock distribution. Specifically, we find that coherent radiation is emitted from two distinct belts nearly parallel to the trench. The position of these belts is in good agreement with the location of the aftershocks, which also are arranged in two disconnected zones of the subduction interface at different depths, the deeper of which is characterised by a large number of repeating event clusters (Rietbrock, Jenkins et al., this session). Thus, our backprojection analysis in combination with the aftershock distribution demonstrates the existence of a peculiar doubled downdip transition from seismogenic behaviour to stable sliding. We suspect fluids released from the downgoing plate to be the cause of the transitions in frictional behaviour because of (1) the co-location of high Vp/Vs ratios with the deep interface seismicity, (2) systematic decrease of depth of onset of deeper seismicity with younging incoming plate age, (3) patchy occurrence along-strike of deeper seismicity.