An efficient multi-scale waveform inversion method in Laplace-Fourier domain

摘要

Aiming at the problem that large computational resources and long computation time are required for the conventional Laplace-Fourier domain waveform inversion, an efficient multi-scale grid algorithm with variable computed area is proposed, and used in inversion modeling of the Marmousi and Overthrust model. This algorithm can choose a proper grid spacing automatically according to the different frequency, and adjust the depth of computing area according to the Laplace damping constant. This algorithm not only improves inversion efficiency significantly without the loss of inversion precision, but also improves the stability due to the decrease of grid number. Inversion results of the Marmousi and Overthrust model demonstrate the validity of the algorithm. In addition, the inversion results by the algorithm still can be approximate to the real model when low frequency information is missing.