Testable geophysical and geological models from probabilistic imaging

Project Overview

Sustainable use of resources in the subsurface require us to be able to manage the risks—financial and otherwise—associated with this process. However, most existing methods to image the subsurface and infer ground physical properties do not enable the best decision-making to take place because they do not include fair estimates of the uncertainties in the physical parameters recovered. In this project, we will build upon existing methods to image the subsurface with unbiased uncertainties, improve them, and use them to establish and answer testable hypotheses about the subsurface. In particular, we will examine the relationship between seismic anisotropy, faulting and how heat is stored in the subsurface using high-enthalpy sites in Ethiopia and Iceland.

The key to our approach is trying many millions of different images of the subsurface properties, and using the collection of those that fit the data best to understand the likely range of properties that could be present at any one point in space.  We can also incorporate our own expert knowledge and other data as prior information using Bayes theorem—a mathematical technique for using what we do know to better understand what we don't.  Using these approaches, we image the subsurface with realistic error bars, and will be able to give operators of subsurface resources uncertainty bounds.  This enables risk-based decision-making and hypothesis testing.

Thus far we have looked at seismic data, which can constrain fractures via seismic anisotropy, which is where seismic waves travel at different speeds in different directions.  However, the imaging methods we develop are well-suited to any type of data, and incorporating other observables such as the electromagnetic field is one possible future direction.