Waveform Q tomography with central-frequency shifts
Subsurface Q (quality factor) structures can provide important constraints for characterizing hydrocarbon reservoirs and interpretating tectonic structures and evolution of Earth in exploration and earthquake seismology. The damping effects of attenuation on seismic amplitudes and phases can be modeled based on the generalized standard linear solid (GSLS) rheology. Compared to traditional ray-based methods, full-waveform adjoint tomography is promising to provide more accurate Q models by numerically solving the viscoelastodynamic wave equation. However, the progress of adjoint Q tomography is impeded by the difficulties of calculating Q sensitivity kernels and significant velocity-Q trade-offs. In this study, following the adjoint-state method, I derive the Q (P- and S-wave quality factors QP and QS) sensitivity kernels, which can be efficiently constructed with memory strain variables. A new central-frequency difference misfit function is designed to reduce the trade-off artifacts caused by velocity errors for adjoint Q tomography. Compared to traditional waveform-difference misfit function, this new misfit function is less sensitive to velocity variations and thus can invert for the Q models more stably suffering fewer trade-off uncertainties.