Improved seismic envelope full waveform inversion by adjoint source analysis

Seismic envelope full waveform inversion (EI) can recover long-wavelength components of subsurface physical properties by fitting of the envelopes (energy) of observed seismic data to the modeled data. For the conventional EI method, the sensitivity of the data envelope perturbations to the velocity model is poor, especially the envelope of reflected waves. One reason for this poor sensitivity is the linearization in the derivation of the gradient, used to update subsurface models. However, after analyzing the adjoint source of EI, we attribute the poor sensitivity problem to the modeled data in the adjoint source, which mutes or degrades the reflection phases as a data-weighting operator. Our improved EI method avoids this issue, is more stable, and increases the resolution in the velocity model. We show tests on the SEAM data, and an Australian marine data example.