Multiscale wave-equation reflection traveltime inversion based on the time-lag method

Abstract As a high accuracy velocity reconstructed method, full waveform inversion (FWI) has been widely applied in geophysical exploration due to the full use of seismic wave information. A notable challenge associated with FWI is the poor precision of initial velocity model. FWI with constrained offset makes it difficult for diving wave to reach deep layers. Reflection wave traveltime inversion can use the wave path information of reflected wave to build a gradient, which has advantages in building a deep-background velocity field for FWI. However, single-time window size can cause a mismatch in traveltime shift extraction. In this study, a time-lag based wave-equation reflection traveltime inversion (TLWERTI) method by frequency division is developed to deal with this issue. First, the gradients between time-lag FWI (TLFWI) and TLWERTI are analyzed to indicate the necessity of time-lag strategy on reflection wave traveltime inversion. Second, a Sigsbee2a model is used to obtain a background model by TLWERTI. The comparison between the results of background inversion model and linear model in FWI highlights the advantage of this method. Third, to test the robustness of TLWERTI in recovering a background velocity field using low signal-to-noise ratio data, a Marmousi noisy data is simulated. Numerical results show that our research method can obtain more accurate initial velocity for FWI, and make the reconstruction process of mid- to deep-layer background velocity more robust and controllable.