The Adjoint state inversion of tsunami source

We develop a novel adjoint-state full waveform inversion procedure to recover the initial water

elevation of a tsunami event. Traditional source inversion using tsunami waves is based on the finite-fault slip modeling. Such inversion methods suffer from the uncertainty of fault parameters or crustal rigidity. Moreover, the heavy computational burden of calculating Green’s functions results in limited spatial resolution and hinders the real-time applicability of the traditional methods to tsunami early warning. In this work, we transplant a popular adjoint full waveform inversion method from exploration seismology to tsunami source inversion. The benefits of the adjoint inversion are three folds: 1) independence of fault parameters, 2) high spatial resolution, and 3) high computational efficiency, especially for dense tsunami arrays and high resolution grids. We valid this approach with synthetic tsunami sources, and apply it to the 2014 Mw 8.1 Iquique tsunami event. Both synthetic and real-data preliminary results show that the adjoint-state method is of high resolution and produces little artifacts, outperforming the traditional tsunami source inversions.