Abstract

Full-waveform inversion (FWI) seeks to find a subsurface model that minimizes the discrepancy between observed seismic data and modeled data. However, the commonly used L2-norm waveform difference misfit functional is susceptible to cycleskipping, primarily due to the existence of local minima. To address this issue, we propose the utilization of the soft-dynamic time warping (SDTW) divergence distance as a more effective misfit metric in FWI. Our approach introduces a hyper-parameter in the formulation, which renders the functional differentiable. This enables the use of the adjoint state method to compute the gradient required for FWI. In contrast to the conventional SDTW, the divergence form of our proposed metric is always positive. The minimum value of the metric is achieved only when the modelled trace closely matches the observed trace. The application of the SDTW-divergence misfit in FWI framework to synthetic and field datasets demonstrates its robustness to cycle-skipping compared to the L2 norm.