Tomographic Velocity Model Building Using Iterative Eigendecomposition

Tomographic velocity model building has become an industry standard for depth migration. However, regularizing tomography still remains a subjective virtue, if not black magic. Singular value decomposition (SVD) of a tomographic operator or, similarly, eigendecomposition of corresponding normal equations, are well known as a useful framework for analysis of most significant dependencies between model and data. However, application of this approach in velocity model building has been limited, primarily because of the perception that it is computationally prohibitively expensive for the contemporary actual tomographic problems. In this paper, we demonstrate that iterative decomposition for such problems is practical with the computer clusters available today, and as a result, this allows us to efficiently optimize regularization and conduct uncertainty and resolution analysis.