Quantifying the incompleteness of the physics model in seismic inversion

Kristopher A. Innanen

At each stage of seismic processing we adopt an underlying physical or mathematical model, and rely on its degree of incompleteness (relative to what is actually experienced by a seismic wave) being either negligible, or irrelevant to the processing task in question. For some tasks we can proceed with remarkably simplistic approximations, but in seismic inversion, our choices are few: any parameter or process affecting the seismic wave has to be included. The geophysical literature reflects the broad range of possible responses to this issue: suppressing elastic conversions in order to use acoustic physics; selecting analysis domains which boost P-wave influences over S-wave, etc. A smaller number of researchers have grappled directly with incompleteness, through probabilistic and numerical schemes. We may be able to come to grips with this problem more deterministically, if we learn to view complex physical models as being perturbations of simple physical models. This will be an important part of a meaningful attempt to apply full waveform inversion / IMMI to multicomponent land data.