GNGTS 2024 - Atti del 42° Convegno Nazionale

Session 3.1 GNGTS 2024 detailed informaton represented by samples in the prior distributon facilitates the reconstructon of a geologically consistent structure of the peat unit, whereas the deterministc method is forced to provide (vertcally) smooth soluton. It is worth notng that B26* is the most reliable borehole as, unlike the others, its core has been extracted and analyzed (Poto et al., 2013), proving the presence of 6 meters of peat followed by roughly 1 meter of gytja (more decomposed peat). Fig. 3 – Vertcal profle presentng the inversion outcomes derived from the Frequency-Domain Electromagnetc Inducton (FDEM) data collected across the Danta peatland in Italy. In terms of inversion performance, the probabilistc approach, relying on forward calculatons of prior samples, proves computatonally efcient and parallelizable. The calculaton of conditonal probability for the entre dataset takes around one hour, while generatng prior samples and associated measurements adds another hour. In contrast, a deterministc inversion of the entre dataset was computatonally infeasible as it required more than a day for the inversion of the 2D secton considered here. Conclusion This research demonstrates the efcacy of a probabilistc petrophysical inversion approach in characterizing the complex subsurface structures of alpine peatlands. By applying ground-based frequency-domain electromagnetc inducton (FDEM) technology to an Italian Dolomite peatland, where peat overlays a conductve clay substrate, the study surpasses the limitatons of traditonal deterministc Occam's inversions. The probabilistc approach, rooted in Bayes' theorem, allows for the incorporaton of realistc prior informaton, ofering a more geologically meaningful representaton of the subsurface. The spatal coherence observed in the results highlights the