Item

Abstract

The surface displacement measurements afforded in the new era of satellite geodesy have proven to be a valuable complimentary data set in addition to the seismic monitoring of active subduction margins. Being able to accurately measure the respective plate motions with high spatial and temporal frequency has inspired many modelling initiatives investigating both heterogeneous plate-interface kinematics and subduction zone seismic cycle dynamics. The isolation of plate interface kinematics from the surface signal and the robust modelling of the kinematic source provide the case-studies against which mechanical models of earthquake recurrence can be benchmarked, and is therefore of utmost importance for intermediate-long term hazard assessment of a subduction margin. In this PhD thesis I present the investigations into the subduction zone seismic cycle plate-interface kinematics and viscoelastic dynamics of the subduction zone in response to the megathrust earthquake. I compare kinematic models to seismicity in order to gauge the heterogeneity in seismic efficiency across the plate interface in both the interseismic and postseismic phases of the earthquake cycle. I attempt to separate the various signals coming from simultaneous postseismic processes and explore and present a discussion of the non-uniqueness of the solution parameter space for the separated signals. For these investigations I use the GPS data and published seismic catalogues from the Chile Maule 2010 Mw 8.8 and Pisagua-Iquique 2014 Mw 8.1 megathrust events. From the Maule investigations, I find that afterslip is a dominant early postseismic process (decaying over 3-4 years) that is well captured due to the excellent coverage of near-field continuous GPS (cGPS). Afterslip spatiotemporal features are well resolved in certain regions and the comparison of afterslip to coseismic slip can reveal regions of the plate interface that are more likely to fail with large magnitude aftershocks. Postseismic processes can be separated if we make some assumptions about afterslip behaviour: To separate the simultaneous postseismic signals I develop the Postseismic Straightening method. The separation considers three postseismic processes: plate interface re-locking, afterslip, and viscoelastic relaxation. Plate re-locking is a traditionally neglected postseismic process that when modelled in combination with viscoelastic relaxation and afterslip significantly improves the model prediction fits to the time series. The effect of re-locking, when added to afterslip and viscoelastic relaxation, is to cause the horizontal displacements to veer with time. The afterslip separated using the Postseismic Straightening method has a normalized decay time function that is in good agreement with the normalized decay time function of the aftershocks over the first 3-4 years, suggesting that the spatiotemporal relation between afterslip and aftershocks persists long into the postseismic time period. The afterslip, however, is predominantly aseismic and pulsing in nature; leading to the interpretation of afterslip pulses on a postseismically weakened plate-interface being triggered by the larger magnitude aftershocks, and with the stress release of afterslip feeding back into the shear stress forcing of the aftershock sequence. For the Pisagua-Iquique earthquake, I investigate the seismic efficiency of the preceding foreshock swarm and the source parameters of the largest foreshock that initiated this swarm. The cGPS motions leading up to the Mw 8.1 mainshock are mainly explained by seismic slip, although significant aseismic postseismic responses can be resolved for the larger earthquakes in the swarm. Similar to the interpretation of the Maule afterslip and aftershocks, I interpret the transient aseismic signals as being afterslip of the foreshocks. While promising spatial relations exist between inter-, co-, and postseismic elastic dislocation kinematics, the degrees of freedom for the slip azimuth (rake) need to be further investigated, especially for the interseismic locking models which can be very sensitive to a constrained backslip azimuth.