Reconstructing near-surface model via joint random-objective waveform inversion of Rayleigh and Love waves

It is getting increasingly popular to use full-waveform inversion (FWI) to estimate near-surface model parameters. Surface waves, including Rayleigh and Love waves, are widely used in shallow seismic, and the joint FWI of both wave types improves the inversion result. Here, we present a random-objective waveform inversion method (ROWI) to invert the waveforms of Love and Rayleigh waves jointly. In the joint ROWI, we build a multi-objective function that accounts for the misfits in the waveform, waveform envelope, and dispersion image of every single shot of Rayleigh and Love waves independently. A stochastic gradient descent method is used to solve the optimization problem. We use both numerical and field-data examples to validate the robustness and efficiency of the joint ROWI method, as well as its outperformance over the conventional joint FWI.