Conclusions

Non-stationary PEFs are an effective interpolator, especially when the model has discontinuities, for missing data problems and an effective regularizer for tomography problems. The weakness of the approach, the requirement that we have field to estimate the PEF from, can be turned into a strength by allowing a geologist's conception of the velocity to be directly encoded into the regularization. A more illuminating test of the non-stationary regularization of tomography would be on a model with sharp, rather than smooth, boundaries.