Seismic wave simulation, full waveform imaging and reverse-time Frechét kernels
Institute of Seismology, National Chung Cheng University, Taiwan
Full wavefield inversion for direct imaging of compressional wave and out-of-plane standing wave velocity distribution is developed, tested and implemented. The inversion strategy is operated in the time-distance domain and is fully tested on wide-aperture, common-shot data. The computational kernel fully utilizes the reverse-time image reconstruction principles. No travel-time picking and phase identification are necessary for full waveform inversion. For each shot records, gradients of misfit function (Frechét derivative) are dynamically determined by cross-correlation between the synthetic (forward) and residual (backward) waveforms. The velocity model is updated by using weighted Frechét derivatives. Convergence to local minima can be avoided by gradually increasing the wavenumber bandwidth in the estimated velocity distribution and to increase the inversion resolution as iterations proceed. Synthetic tests show that the effects of the multiples, scattering, artificial boundary reflections, physical or numerical noise do not contaminate the final results. Stable convergence is guaranteed to the correct solution. Using full two-way waveform approach for seismic imaging simplifies un-necessary skeleton seismic processing procedures. However, the resolution is limited by the bandwidth of seismic data, source wavelet and dominant frequency. Convergence rate and stability of our in-house developed inversion algorithm is depends on step length and the complexity of subsurface structure associate with the steepest decent direction. For land data, near-surface effects including topography, lateral velocity variation, source and receiver static corrections are automatically included. For marine seismic data, multiples generated by water layer can be effectively suppressed through wavefield based seismic processing approach. We illustrate the effects on the construction of reverse-time Frechét derivatives for different type source-receiver configuration, velocity variation and topography changes. The images of Frechét kernel provide a unique opportunity in understanding the characteristics and limitations of reverse-time imaging strategy for different types of data gathers.