Potency density tensor inversion of complex body waveforms with time-adaptive smoothing constraint

Abstract

SUMMARY Large earthquakes are often accompanied by complex fault rupture, but it has been difficult to reliably estimate such a complex rupture process with conventional waveform analysis tools due to modelling errors originating from limited accuracy of the fault geometry. Recently, a potency density tensor inversion method has been developed to solve this problem; allowing any types of faulting mechanism on an assumed model plane, the method replaces the modelling error of fault orientation with that of fault location, which is insensitive in teleseismic waveforms with low spatial resolution. The method has successfully unveiled earthquake source processes with geometrically complex fault rupture. However, the method imposes the same intensity of smoothing constraint on all the basis slip components irrespective of possible changes of slip direction during the rupture. This leads to excessive smoothing to a slip component with large amplitude, which results in obscuring the rupture process. In this study, we propose a time-adaptive smoothing constraint that dynamically adjusts the smoothness intensity inversely proportional to the amplitude for each basis slip function. Through a numerical experiment assigning an input model involving a drastic change in the focal mechanism (reverse, strike-slip and normal faulting) during the rupture, we find that the time-adaptive smoothing constraint solves the problem of excessive smoothing to the dominant slip component, and the spatiotemporally non-uniform rupture episodes with different focal mechanisms are successfully reproduced. To evaluate the feasibility and effectiveness of the time-adaptive smoothing constraint, we apply the method to the teleseismic body waves of the 2002 Denali fault and the 2008 Wenchuan earthquakes, which involve complex fault ruptures with changing focal mechanisms. We find that the developed method well captures the focal mechanism transition in space and time from reverse to strike-slip faulting during the ruptures of the 2002 Denali fault and the 2008 Wenchuan earthquakes. Even though these source models are built using only the teleseismic P waveforms with simple model fault geometry that is represented by a horizontal rectangular plane, they well explain the complex observed waveforms and agree with characteristics of source processes obtained in previous studies using seismic and geodetic data as well as field surveys. The potency density tensor inversion method with time-adaptive smoothing constraint is a powerful tool to analyse earthquake rupture processes with complex fault geometries involving different faulting types.