Full waveform inversion using the PSPI migration: a convergence study

Marcelo Guarido, Laurence R. Lines, Robert James Ferguson

A full waveform inversion (FWI) routine using PSPI migration with a deconvolution imaging condition was tested on an acoustic synthetic 2D survey using the Marmousi model and the inverted P-wave velocities are, in the worst case, promising. The model update is computed by averaging monochromatic scaled gradients at each iteration and the resulting model is an improvement of the one obtained in the previous work (Guarido et al., 2014). We tested different starting models to check the routine behavior and, as expected, better the initial model, better will be the migration of the residuals and the model update, resulting in higher resolution velocity inversion. The conjugate gradient was included in the routine, building more precise gradients and increasing the quality of the inverted model. Impedance inversion by trace integration was applied in the model update (gradient) showing promising results, mostly related to the inversion of thicker layers, but it still requires improvement.