3D Monte Carlo Inversion of Surface Magnetic Resonance Measurements

Magnetic Resonance Tomography (MRT) is the only geophysical method where the measured signal is directly related to water distribution in the ground. This property allows three-dimensional imagery of water content by signal inversion routines. Because of its non linearity, the inverse problem has a quasi-infinite number of solutions implying as many possible spatial distributions of water content. A good answer to this problem, relevant for ice cavity detection and karstic structures mapping, is to provide a set of solutions consistent with the measured data. Markov Chain Monte Carlo (MCMC) algorithm applied to the MRT inverse problem provides a random exploration of the solutions giving the ability to compute probabilistic answer to a particular data set. For saturated structure detection, first results on synthetic cases demonstrate the routine ability to show an important anisotropy in the MRT resolution. Finally, a real case study, where the MCMC and linear inversion are compared, also shows the benefit of the method for a French Alp glacier cavity detection.