I³MS - Hauser Seminar
Dr. Jürg Hauser - A Pragmatic Probabilistic Approach for the Inversion of AEM Data
Csiro, Perth, Australia
Airborne electromagnetic data is widely used in greenfield exploration for the delineation of features of interest such as near-surface geological interfaces or basement conductors. From a theoretical point of view, probabilistic approaches have numerous advantages when compared to the commonly employed deterministic approaches. Yet for them to be embraced by industry probabilistic approaches need to offer an attractive compromise between efficiency and exhaustive stochastic search in the inference of models. In this talk I will focus on the search for such a probabilistic approach that is ultimately applicable on the tenement scale. I will introduce a Bayesian parametric bootstrap approach as a potential alternative to the widely employed Markov Chain Monte Carlo (MCMC) algorithms. By treating prior information as implied observations, a parametric bootstrap can be used within a Bayesian framework. When inverting for spatially coherent interfaces the proposed bootstrap methodology is likely to explore model space sufficiently, and is computationally more efficient than an MCMC approach. Since the true value of any Bayesian approach lies in predictive applications, I will finish with an example where uncertainties quantified in an inversion of AEM data are propagated into a groundwater model.