Source-independent elastic frequency-controllable envelope inversion

Abstract

Elastic full-waveform inversion (EFWI) is a high-resolution inversion method that minimizes full-waveform misfit between synthetic and observed data under the seismic elastic wave equation. In practice, EFWI effectiveness depends strongly on the accuracy of the source wavelet and the initial models, as errors in either degrade the synthetic-observed match and drive the inversion toward local minima. To mitigate these issues, we develop a novel source-independent initial model building method for EFWI, termed source-independent elastic frequency-controllable envelope inversion (SIEFCEI), which enables accurate recovery of low-wavenumber P- and S-wave velocity models from multi-scale elastic frequency-envelope data without relying on an accurate source wavelet. The misfit function of SIEFCEI is formulated based on an amplitude-semblance function to eliminate the need for an accurate source wavelet during inversion, where the matching data in the misfit function are computed using the elastic frequency- controllable envelope operator. SIEFCEI exhibits two notable advantages. First, the data used in the inversion contain richer low-wavenumber information than conventional envelope data, owing to the frequency-controllable mechanism, which reduces the risk of convergence to local minima. Second, SIEFCEI avoids the need to select an optimized reference trace, as required by traditional source-independent approaches, which improves robustness under inconsistent source wavelets. Numerical tests on both synthetic and field data demonstrate the effectiveness of SIEFCEI when starting from poor initial models and using an inconsistent source wavelet.