Multisource waveform inversion of marine streamer data using normalized wavefield

Abstract

Multisource full-waveform inversion based on the [Formula: see text]- and [Formula: see text]-norm objective functions cannot be applied to marine streamer data because it does not take into account the unmatched acquisition geometries between the observed and modeled data. To apply multisource full-waveform inversion to marine streamer data, we construct the [Formula: see text]- and [Formula: see text]-norm objective functions using the normalized wavefield. The new residual seismograms obtained from the [Formula: see text]- and [Formula: see text]-norms using the normalized wavefield mitigate the problem of unmatched acquisition geometries, which enables multisource full-waveform inversion to work with marine streamer data. In the new approaches using the normalized wavefield, we used the back-propagation algorithm based on the adjoint-state technique to efficiently calculate the gradients of the objective functions. Numerical examples showed that multisource full-waveform inversion using the normalized wavefield yields much better convergence for marine streamer data than conventional approaches.