Crosshole acoustic velocity imaging with full-waveform spectral data: 2.5-D numerical simulations

Abstract

This paper focuses on the question of full-waveform inversion and the use of crosshole full-waveform spectral data. We examine the differences in effectiveness of each form of data (real, imaginary, amplitude, phase and Hartley spectra) and determine which is best for imaging the velocity distribution. By 2.5-D numerical simulation for three models, we found that, except for the phase data, the monochromatic real, imaginary, amplitude and Hartley spectra can be used to image the targets between boreholes by full-waveform inversion; the real and imaginary spectra produce nearly the same quality of images. The images obtained from the amplitude data exhibit more artefacts than the others. The inversion of the Hartley spectral data gives the best image of all. By computing the data misfit functions, we found that the profile of the misfit function of the phase data is more complicated than the others. For example, some discontinuities or fluctuations occur in the neighbourhood of the true solutions. The complexity of the misfit function may be the main cause for the failure to satisfactorily image with the phase data alone: local minima capture the misfit function during the attempted global optimisation.