Discrepant stress distributions around instability regions: A new view for earthquake nucleation zones prediction

Abstract

It is believed that seismic failure conditions are sensitive to strain-softening behavior of nominal rock or fault gouge, and that precursors prior to a big earthquake (i.e., tectonic trains, water level changes, and Vp/Vs anomalies, etc) are provided by the acceleration of local slip. Previous studies of earthquake nucleation on laboratory faults show that the initiation of unstable fault slip is spatiotemporal dependent and consists of an interval of fault preslip (or creep) that localizes and accelerates to a dynamically propagating rupture. We pose that perturbation-type experiments can provide a natural condition to help analyze the potential mechanisms between instability events and the stress loading. In this study, we conducted three sets of double-direct shear experiments on a 300 mm long fault filled with gypsum-rich gouges, under normal stress of 10 MPa superimposed with perturbations of various amplitudes (i.e., 0-0.5 MPa) and a fixed frequency (0.1 Hz). The result showed that during each cycle of the stick slip behavior, the applied normal stress perturbations were redistributed along the fault zone as revealed by the along-fault strain measurement in the normal direction. As such, the fault can be divided into different zones characterized by varied coupling with respect to the applied perturbations. We found, coincidently, nucleation of the final instability, as revealed by the strain measurement in the shear direction, tended to occur at the boundary between the so-called strong and weak coupling zones (‘Transition Zone). Moreover, local normal stress near the nucleation zone also showed some weakening prior to the instability, which was similar to that seen in the local shear stress, and hereafter referred to as ‘normal failure’. Based on these observations, we proposed an empirical equation to fit the normal strain or stress data, giving the distribution of the coupling coefficient (c-value) and the anomaly (a-value) along the simulated fault. Finally, we applied the proposed equation to fit the water level data from 6 monitoring stations along the fault that hosted a nature earthquake (~ML 4). The fitting results predicted a Transition Zone, which was close to the hypocenter. In the end, we propose that this approach can be tested widely to natural observations of various precursory signals, especially those considered to be sensitive to fault-normal deformation (“dilatation” or “compaction”), such as water level, soil gas, and Vp/Vs anomalies. Hopefully, the results can shed some lights on the location of the earthquake nucleation zone.