GNGTS 2021 - Atti del 39° Convegno Nazionale

401 GNGTS 2021 S essione 3.1 2D STOCHASTIC INVERSE MODELING AND 3D INVERSION OF MAGNETOTELLURIC DATA FROM THE LARDERELLO-TRAVALE GEOTHERMAL FIELD F. Pace 1 , A. Santilano 2 , A. Manzella 2 , A. Godio 1 1 Department of Environment, Land and Infrastructure Engineering (DIATI), Politecnico di Torino, Torino, Italy 2 Institute of Geosciences and Earth Resources - National Research Council (IGG-CNR), Pisa, Italy Introduction The Magnetotelluric (MT) method is one of the most effective geophysical techniques for investigating deep geothermal systems because it can image the electrical-resistivity distribution of the Earth from a few meters to hundreds of kilometers of depth (Chave and Jones, 2012). The MT inverse problem is ill-posed in nature with nonlinear and equivalent solutions. The standard approach to solve the inverse problem is the iterated and linearized inversion. It is also possible to perform stochastic inverse modeling by adopting Monte-Carlo or metaheuristic methods. Global search methods have become of major interest in geophysics because they are theoretically able to find the global minimum of a function as the final solution without being trapped in one of the several local minima. The potential advantages of metaheuristics are also to ensure a complete sampling of the search space of solutions and the independence from the startingmodel. Stochastic inverse modeling of MT data from geothermal areas has considerable potential, mainly in those cases where the geological complexity and the difficulty in retrieving reliable external constraints can negatively affect the solution of the inverse problem. One of themost extraordinary geothermal resources globally is located in the Larderello-Travale geothermal area (LTGA), Italy. The site has been the object of extensive industrial and scientific research over the past century. Nonetheless, some geological, physical and chemical aspects are still a matter of investigation. This work presents recent advances in the interpretation of MT data from the LTGA. The data sets acquired during the last decades were re-examined following the most recent techniques of MT data analysis of the impedance and tipper tensors. The geoelectrical dimensionality, directionality and phase tensor properties were determined. Then, two MT profiles from LTGA were interpreted by using 2D stochastic inverse modeling, namely, the Particle Swarm Optimization (PSO) algorithm. Finally, we focused on the Travale geothermal field to apply the state of the art of MT inversion techniques, that is, 3D derivative-based inversion. It is nowadays of pivotal importance to characterize the geoelectrical structures from geothermal areas for the simple and obvious reason that we need 3D inversion methods to model a 3D Earth. Stochastic inverse modeling of MT data by employing Computational Swarm intelligence Computational Swarm Intelligence is a sub-branch of Artificial Intelligence and encompasses several nature-inspired population-based metaheuristic methods. One of these algorithms is PSO, which has been widely applied to geophysical data (for a review, see Pace et al., 2021). The application of PSO to solve the 2D MT inverse problem has been investigated in Pace et al. (2019a), where the algorithm was accurately calibrated to enhance the stability and convergence of the solution and was parallelized to speed up the computation. The method was validated on synthetic data and then applied to field data. PSO provided a resistivity model of the Earth in line with results from previous interpretations and demonstrated that there is no need for a priori initialization to obtain robust 2D models. The stochastic nature of PSO and the combination of exploration and exploitation behaviors played a crucial role in finding the global minimum of the search space as the final solution.