Salt reconstruction with full waveform inversion and a parametric level-set method

Abstract

Seismic full-waveform inversion has become a popular technique for imaging the subsurface from seismic data. The reconstruction of the subsurface medium parameters becomes challenging in the presence of sharp contrasts such as salt bodies. We address the problem by splitting the subsurface model in two parts: a background velocity model and a salt body with known constant velocity but undetermined shape. The salt geometry is represented by a level-set function parametrized by radial basis functions. This leads to a non-linear optimization problem with a modest number of parameters. We improve the convergence speed by utilizing second-order updates from Gauss-Newton Hessian. Tests on the suite of idealized salt models show that the proposed method accurately determines the salt geometry in the presence of a modest amount of noise.