Global inversion and parametrization for building tomographic velocity models

Abstract

Traveltime tomography is applied to investigate seismic structures of the Earth's subsurface. An accurate tomographic velocity model is important for a high-resolution waveform velocity building and its availability is one of the main components to mitigate the nonlinear inverse problem. We present a new methodology of obtaining velocity models for traveltime tomography studies. We found a way to get a highly accurate first-arrival traveltime tomography in combination with global optimization. The role of global optimization is twofold: to find initial solutions that are close to ‘truth’, and to guide tomographic inversion towards a geologically consistent model that explains the data. The main advantage of our workflow is a data-driven approach avoiding the use of a conventional layer-based parameterization and incorporation of manual interpretations into the velocity model. To date, a few geophysical studies have been focused on developing data-driven and a labour non-intensive regional tomographic velocity model building workflow. In our study, we present the tomographic velocity model building workflow as a combination of first-arrival traveltime tomography and global optimization. Global optimization allows to search for velocity parameters and depth to interfaces in the larger search area with a higher chance of convergence. After defining the geometry of main layers and general velocity trends, traveltime tomography with a bi-cubic B-spline model parameterization can be fitted to further update the velocity model. Our approach allows obtaining a highly accurate velocity model which can be used for seismic depth migration and as a starting model for a FWI seismic imaging. The workflow is developed and applied to synthetic and field regional seismic datasets. The developed methodology is applied for a shallow seismic engineering data and regional Ocean Bottom Seismic data. We identify four key components that lead to building an accurate tomographic velocity model: (i) understanding prominent horizons and possible velocity distribution of a layer within the study area. (ii) Performing ray penetration test to define offset ranges which carry the velocity information for the defined layers. (iii) Determining inversion schema to a perform global search for the velocity trends and major boundaries, and a local search to update lateral velocity variation. (iv) Iteratively update a set of defined layers (i.e., sediment, igneous crust and basement) in a top-down manner.