A study of upper mantle anisotropy in Cascadia using teleseismic shear wave delays

Seismic tomography is one of the most powerful tools available when attempting to characterise the Earth’s interior. Often an underlying assumption of these inversions is one of elastic mantle isotropy, when it is known that anisotropy is present within the upper mantle. Simple shear deformation of the mantle creates lattice-preferred orientation (LPO) of minerals, principally olivine which is an anisotropic mineral. Therefore where an LPO forms we expect a seismically anisotropic upper mantle. When a linearly polarised shear wave (S-wave) encounters this anisotropy it splits into two orthogonal waves with one component polarised along the plane of greatest seismic velocity. Determining the direction of anisotropy can reveal information about deformational history, flow and other properties in the upper mantle. Accounting for anisotropy will lead to more accurate seismic tomography models whilst providing indications of convective flow and structural fabrics within the upper mantle. The area of this study is the Cascadia Subduction Zone in the Western USA. The relatively young and small Juan de Fuca (JDF) plate is subducting beneath the North American plate at the Cascadia trench. It is a complex geodynamic setting which comprises: JDF ocean ridge, Cascadia trench, forearc between the trench and Cascadia volcanic chain, a back arc region. This is bound to the North and South by triple junctions. This study utilises 593 seismic stations on 25 different networks and data from September 2011 to August 2015. Previous anisotropic S-wave tomography studies have focused mainly on the SKS phase. Conversion from P to S-wave at the core mantle boundary ensured that only anisotropy within the mantle will influence the S-wave arrival. This study instead focuses on 4 different S-wave phases: S, sS, Sdiff and sSdiff. Teleseismic S-wave arrivals for 383 earthquake events were identified on the transverse channel of the seismic station. Initially the S-wave arrivals were analysed with reference to the AK135 earth reference model. Alignment of the S-waves was undertaken using multichannel cross correlation (MCC). A stacked version of all the S-wave arrivals was created to identify the waveform. The measured delay times and the stacked traces arrival time were combined to calculate the travel time. The stacked traces were then rotated to find the azimuth of the polarisation angle, the angle of maximum S-wave energy arrival. Subsequently a second MCC was completed in the polarisation direction to determine the delay time of the polarised wave, containing information about both isotropic and anisotropic velocity variations along the ray path. Inverse tomographic modelling was completed using the updated travel times from both the transverse channel and the polarisation channel. Initially a deterministic inversion was utilised to produce 700km deep, 3D S-wave velocity models. For anisotropic inversions, a vector aligned with the axis of symmetry of the hexagonal velocity model was also calculated, allowing a visual representation of anisotropy. Following these results, a set of stochastic inversions based on the reversible jump Markov Chain Monte Carlo (rjMCMC) methodology were completed with the intention of comparing the two inversion methods for the same dataset. From initial interpretations from both the isotropic and anisotropic deterministic inversion results, the subducting JDF plate can be observed as a steeply dipping fast velocity anomaly. Both models also display evidence of a possible gap within the subducting plate. A roughly horizontal circular pattern in the anisotropy vector at the southern edge of the JDF slab could be interpreted as toroidal mantle return flow caused by slab rollback of the subducting plate. Initial results from the isotropic stochastic inversion are comparable to the deterministic inversions. A stochastic anisotropic inversion is currently being undertaken.