Towards a Refined 3D Model of the Western Pacific Slab to Investigate the Nature of Deep Earthquakes

The nature of deep earthquakes with depths greater than 70 km is enigmatic because brittle failure at this high-temperature and high-pressure regime should be inhibited. The existing seismological constraints can't yet definitively distinguish between the three major hypotheses of what causes deep intraslab earthquakes, dehydration embrittlement, phases transformational faulting, and thermal runaway instability. One of the main reasons is that the fine 3D structures of the subducting slabs are not well constrained, e.g., slab upper interface, thickness, and internal fine layering. This study aims at developing a refined 3D model of the Western Pacific slab in Kuril, Japan and Izu-Bonin based on a large seismic waveform data set and full waveform inversion. Our data set contains three-component waveforms of about 150 earthquakes recorded on over 2,000 broadband stations mostly located in China, Korea, Japan, and Mongolia. The full waveform inversion (FWI) of this study will update the model by iteratively minimizing the misfit of waveform shape and phase between the data and synthetics based on spectral-element and adjoint methods. The radially anisotropic model of the Western Pacific is updated using the 3D finite-frequency kernels and a conjugate gradient method. We start the FWI with minimizing the misfit in the intermediate period ranges of 20 - 120 s for body waves and 40 - 120 s for surface waves and gradually incorporate the misfit in the shorter periods down to 10 s for body waves and to 20 s for surface waves. Our initial model is composed of two previous models in East Asia, models EARA2014 and FWEA18, a 3D crustal model, CRUST1.0, and a smooth 3D mantle model, S362ANI. The earthquake centroid depths are relocated with grid search within the initial model and the source parameters are subsequently reinverted with 3D Green's functions. Source parameter inversion are performed after every five iterations of the structural model. Here we present a preliminary model of the Western Pacific slab. The ultimate goal of this study is to use the higher resolution slab model to constrain the spatial relationships between the slab structural details and the deep earthquakes and therefore, better understand the nature of deep earthquake.