UC San Diego

This dissertation is an attempt to apply the powerful tool of Bayesian inference to the elegant physics of controlled source electromagnetic (CSEM) propagation through marine sediments with high resistivity contrasts. CSEM is highly effective in the detection of resistive hydrocarbon accumulations in conductive sediments, but its use requires careful analysis owing to the problems of sparsely sampled data, noisy observations and non- uniqueness. Bayesian inference allows us to tackle these limitations in a quantitative, probabilistic framework by using prior information about the geology in question and the statistics of the data noise. We extend conventional Bayesian analyses to a second level, where we infer the complexity of subsurface resistivity models required to explain the observed noisy CSEM data, from the observations themselves. Although we have focused on the CSEM method, the techniques discussed are generally applicable to many problems in the geosciences