Deep learning Laplace-Fourier full-waveform inversion with virtual supershot gathers

In the past decade, full waveform inversion (FWI) has become a powerful tool to obtain high-resolution subsurface velocity and imaging. The conventional way to solve this ill- posed inverse problem is through iterative algorithms. However, it suffers issues like high computational cost and cycle-skipping. Additionally, the application of FWI to land seismic data faces difficulties related to the complex physics, unknown and spatially varying source signatures, and low signal-to-noise ratio (SNR) in the data. In this study, we propose a new method to reconstruct the velocity model from seismic data organized in the virtual super gathers (VSG) in the Laplace-Fourier domain by deep learning network (DNN). The proposed new method is applied to a synthetic data experiment, which demonstrates that the application of DNN on VSG in Laplace-Fourier domain provides an effective solution for high-resolution velocity mapping in complex near-surface conditions.