摘要

We propose a strategy to overcome the high sensitivity to early-time noise of the Laplace-domain waveform inversion. In deep-sea seismic data, this problem is particularly crucial to obtaining accurate velocity structures. To this end, rather than simply filtering or muting early-time data, we propose replacing the original, noise-contaminated direct waves with analytically computed noise-free waves. To reconstruct the noise-free direct waves, we compute Green's functions for half-space media, estimate the source wavelet from the original direct waves using the full Newton method in the frequency domain and then convolve the Green's functions with the estimated source wavelet. The data obtained by merging the reconstructed direct waves with the original late-time data set can then be used for Laplace- and Laplace-Fourier-domain waveform inversions. To verify the source estimation and direct wave reconstruction strategy, we applied it to field data acquired in a deep-sea environment and obtained a realistic 2-D velocity model. The source estimation and direct wave reconstruction methods can also be applied to 3-D Laplace-domain waveform inversion.

  • 出版日期2011-11