Source-independent elastic envelope inversion using the convolution method

Abstract

Elastic full waveform inversion (EFWI) is a powerful technique. However, its strong non-linearity makes it susceptible to converging towards local extremes during the iterative process due to various factors like insufficient low-frequency information or an inadequate initial model. The existing elastic envelope inversion can offer a promising initial model for EFWI when low-frequency information is unavailable, reducing the dependence on both the initial model and low-frequency data. However, its accuracy is affected by the quality of the source wavelet, potentially causing the EFWI to run in the wrong direction if there is a discrepancy between the simulated wavelet and the field wavelet. To address these issues and enhance the reconstruction of large-scale information in the model, we propose a novel approach called source-independent elastic envelope inversion, employing the convolution method. By combining this method with source-independent multiscale EFWI, we effectively establish P- and S-wave velocity models even in situations with inaccurate wavelet information. The results of testing on a portion of the Marmousi2 model demonstrate the effectiveness of this technique for both full-band and low-frequency missing data scenarios.