Maximum entropy reflection tomography

Abstract

Conventional reflection tomography attempts to reconstruct an image of the subsurface velocity field by minimising the differences between measured travel times and the travel times through the proposed image. Such schemes have been widely documented. Conventional algorithms also parameterise depths to major reflectors in the inversion, and hence necessitate that data times be measured consistently from these reflectors. In areas of complex geology, strong coherent reflectors may be hard to find. In such cases, conventional tomographic imaging would be at least difficult to work with, and may be completely inapplicable. The angles of emission and reception can be obtained from the standard reflection seismic data to enable ray tracing for travel times without reflector parameterisation. This means any moderately strong event can be included in the inversion and allows simpler automatic data picking routines to be employed. Any inversion that is controlled by a priori and/or smoothness constraints implicitly selects a solution that fits the user's preconceptions to some degree. The maximum entropy image, on the other hand, ensures that the solution contains no structure other than that implied in the data. Inversions controlled by a maximum entropy criterion have already been used successfully in astronomy and medical imaging and have similar application in seismic traveltime inversion. To facilitate rapid convergence, this non-linear maximum entropy inversion incorporates subspace searching methods and stages of decreasing parameterisation scale lengths.