Velocity inversion by coherency optimization

: We introduce an approach to velocity and reflectivity estimation based on optimizing the coherence of multiple shot-gathering inversions of reflection seismograms. The resulting algorithm appears to avoid severe convergence difficulties reported for output (nonlinear) least- squares inversion. We describe in detail an algorithm appropriate for plane-layered acoustic models, using the convolutional approximation to the plane-wave (p-tau) seismogram. We give theoretical and numerical evidence that coherency optimization, as defined here, yields stable and reasonably accurate estimates of both velocity trend and reflectivity, by exploiting reflection phase moveout and amplitudes in a computationally efficient way. We demonstrate that the approach may be applied to field data by extracting velocity and reflectivity estimates from a Gulf of Mexico marine data set. Finally we explain briefly how the approach may be modified to determine elastic models and source parameters as well as to determine laterally heterogeneous models.