Tomographic velocity model building in high-fidelity seismic depth imaging of near surface and VSP settings

The inversion nonuniqueness and artifacts of traveltime tomography can be resulted from the mismatch between the raypath distribution and the model parameterization scheme. A reasonable model parametrization is the key to successful tomography applications, especially when the ray coverage is poor and uneven. Motivated by a geologically plausible model parameterization, the multiscale deformable layer tomography (DLT) has been developed to estimate the velocity field that can be approximated by some velocity layers. Based on the previous work of DLT, I improve and extend the method to tackle some challenging velocity model building (VMB) tasks for high-fidelity depth imaging purposes, which aims at imaging the targeted features at their correct positions. The first VMB challenge I investigate is about complex near-surface with strong variations in surface topography and weathering velocities. I propose a layer-cell tomography that sequentially implements DLT and cell tomography. The idea is to use DLT to invert for the long-wavelength variations first, and then use cell tomography to invert for the short-wavelength variations within the resolved layered model. The benefit of this layer-cell approach is to deliver a robust and accurate solution model while suppressing the smearing artifacts where angular ray converge is poor. Both the synthetic and field data examples support the effectiveness of the method to obtain high-fidelity depth images of deep structures. I explore the potential of DLT to estimate velocities in imaging salt domes, particularly focusing on the salt-related converted waves acquired under a vertical-seismic-profiling (VSP) geometry. I develop a joint DLT that utilizes both the P-wave and converted-wave traveltimes to invert for the salt flank geometry. In a physical model test, the joint DLT delineates the flank geometry of the salt model accurately. In addition, I investigate a single-offset VSP data acquired over a salt dome in southern Texas, in which the salt flank image is contaminated by some converted waves generated by a layer of anhydrite over the top-of-salt. Taking the first-arrival traveltimes of the single-offset shot as the input, I apply a constrained multiscale DLT to resolve the lateral variations of the anhydrite layer. The solution model is critical for the identification and adaptive subtraction of the converted-wave contaminations.