Fracture-related wavefield polarization and seismic anisotropy across the Greendale Fault

We investigate seismic signatures of fracturing in a newly ruptured strike-slip fault by determining the wavefield polarization in the New Zealand Canterbury Plains area and across the Greendale Fault, which was responsible for the 3 September 2010 Darfield Mw 7.1 earthquake. Previous studies suggested that fractured rocks in fault damage zones cause directional amplification and ground motion polarization in the fracture-perpendicular direction as an effect of stiffness anisotropy, and cause velocity anisotropy with shear wave velocity larger in the fracture-parallel component. An array of 14 stations was installed following the Darfield earthquake. We assess polarization both in the frequency and time domains through the individual-station horizontal-to-vertical spectral ratio and covariance matrix analysis, respectively, and compare the results to previously reported anisotropy measurements from shear wave splitting. Stations installed in the Canterbury Plains have an amplification peak between 0.1 and 0.3 Hz for both earthquakes and ambient noise. We relate the amplification to the resonance of a considerable thickness (c. 1 km) of soft sediments lying over the metamorphic bedrock. Analysis of seismic events revealed the existence of another peak in amplification between 2 and 5 Hz at two on-fault stations, which was not visible in the noise analysis. In contrast to the lower frequency peak, the ones between 2 and 5 Hz are more strongly anisotropic, attaining amplitudes up to a factor of 4 in the N52° direction. To interpret this effect we model the fracture pattern in the fault damage zone produced by the fault kinematics. We conclude that the horizontal polarization is orthogonal to extensional fractures, which predominate in the shallow layers (<2 km) with an expected strike of N139°. Fracture orientation is consistent with coseismic surface rupture observations, confirming the reliability of the model. S wave splitting is produced by velocity anisotropy in the entire rock volume crossed along the seismic path; thus, it is affected by deeper material than the amplification study. We explain the rotation of S wave fast component observed by Holt et al. (2013) near the fault in terms of the dominant synthetic cleavages at greater depths (>2 km), expected in N101° direction on the basis of the model. Thus, different fracture distribution at different depths may explain different results for amplification compared to anisotropy. We propose polarization amplification analysis as a complementary method to S wave splitting analysis. Polarization analysis is rapidly computed and robust, and it can be applied to either earthquakes or ambient noise recordings, giving useful information about the predominant fracture patterns at various depths.