Near-surface geophysics techniques have proven their reliability in various application fields, such as geotechnical engineering, resource exploration, and archaeological research. Their success in these contexts is closely related to data interpretation methods, which must be able to resolve shallow structures/bodies that are small and/or located at short distances from each other, thus preventing their identification and discrimination due to the overlapping effects generated by different sources. Here, we propose a new data inversion tool, based on a probabilistic Bayesian approach, which is able to scan near-surface magnetised structures. The developed algorithm allows selecting the most-likely probability density function, associated with the most-likely magnetic susceptibility contrast distribution in the explored model space, by refining the discretisation of the anomalous areas, i.e. those areas corresponding to the highest susceptibility contrast. We validated the algorithm on synthetic magnetic data generated by anthropogenic-like bodies, and, then, inverted experimental magnetic measurements acquired at the archaeological site of Phaistos (Crete, Greece). In the latter case, the retrieved most-likely model fits well with the remains found in the study area, and correctly identifies the proper magnetic susceptibility contrast, thickness, and depth of the Minoan wall top brought to light by the archaeological excavation.
An innovative probabilistic Bayesian tool to scan buried magnetised structures: testing on the Phaistos (Greece) archaeological site