Recover low-frequency for full-waveform inversion via band-limited impedance inversion and projection onto convex sets
Wenyong Pan, Kristopher A. H. Innanen, Gary F. Margrave, Scott Keating
Full-waveform inversion (FWI) provides high-resolution estimates of the subsurface properties by iteratively minimizing an l-2 norm misfit function, which measures the difference between the modelled data and observed data. FWI suffers from cycle-skipping difficulty arising from inaccurate initial model and lack of low frequency information in the seismic data. In this paper, we aim at recovering low frequency information from well log data through band-limited impedance inversion. Projection-onto-convex-sets (POCS) algorithms are generally used to infill the missed traces in seismic data reconstruction. In this paper, we also consider to recover the low frequency information by spectral extrapolation with POCS algorithm. The reflectivity estimate is first generated. The frequency spectrum is then extrapolated with POCS algorithm. The band-limited impedance inversion is then performed with the reflectivity section and interpolated well log data. Through, this process, we build an enhanced initial model for full-waveform inversion with low frequency information. We illustrate with numerical examples that the inversion results can be improved with the enhanced initial velocity model.