Sensitivity kernels for
wave-equation migration velocity analysis

Paul Sava and Biondo Biondi

paul@sep.stanford.edu, biondo@sep.stanford.edu

ABSTRACT

The success of migration velocity analysis methods is strongly dependent on the characteristics of the linearized tomographic operator that is inverted to estimate velocity updates. To study the properties of wave-equation migration velocity analysis, we analyze its sensitivity kernels. Sensitivity kernels describe the dependence of data space elements to small changes of model space elements. We show that the sensitivity kernels of wave-equation MVA depend on the frequency content of the recorded data and on the background velocity model. Sensitivity kernels computed assuming the presence of a salt body in the background velocity show that these kernels are drastically different from idealized ``fat rays''. Consequently sensitivity kernels cannot be approximated by artificial fattening of geometrical rays. Furthermore, our examples illustrate the potential of finite-frequency MVA as well as the frequency-dependent nature of illumination for subsalt regions.