Bayesian full-waveform tomography of Ground Penetrating Radar dataPP

We present our newly developed probabilistic full-waveform tomography algorithm to invert crosshole Ground Penetrating Radar data for the subsurface porosity distribution. The algorithm samples the solution space with an efficient Markov chain Monte Carlo sampler using multiple chains and differential evolution of past states to determine the update direction. The dimensionality of the problem is reduced with the circulant embedding technique. The porosity is linked with the electric permittivity and the conductivity through petrophysical relations.

This is poster number 35 in Poster Session

Jürg Hunziker (Université de Lausanne)
Eric Laloy (Belgian Nuclear Research Center)
Niklas Linde (Université de Lausanne)
bayesian methods, computed tomography, inverse problems