Probabilistic multiphysics inference with flexible coupling and data covariance estimation

Abstract

We consider combining complementary information contained in multiple data types recorded from distinct physical processes interacting with the Earth’s subsurface. Such multiphysics inference of non-invasive geophysical observations can improve the resolution of Earth structure and processes, but is plagued by many subjective choices that practitioners are commonly required to make. We present the method of probabilistic multiphysics inference that employs Bayesian statistics to overcome several requirements for subjective choices. To ensure appropriate data weights, full data covariance matrices are estimated during Markov chain Monte Carlo sampling. The layering structure of the subsurface is estimated with flexible coupling, where the number of homogeneous layers is treated as unknown and the number of geophysical parameters for each layer are unknown. The latter permits flexible coupling such that parameters for different physical processes are not required to share the same layering structure, which avoids over-parametrization. We consider two examples with elastic and electromagnetic waves. In the first example, the thicknesses of shallow (tens of meters) active and permafrost layers are better constrained by probabilistic multiphysics inference. The second example resolves cratonic structure, with reduced uncertainty of a sedimentary basin and for the depth of the lithosphere-asthenosphere boundary. [Work supported by an NSERC Discovery Grant.]