Predicting electrical resistivity in natural degassing geological systems through petrophysical and thermodynamic data: a machine learning approach

Natural degassing geological systems have important implications for human safety, ecosystems, groundwater quality, and volcanic hazard. Therefore, their investigation and monitoring are essential for assessing the hazards associated with the upward migration of soil gases. Electrical resistivity is widely recognized as a key parameter for characterizing these systems, as it is sensitive to water saturation, gas content, and fluid temperature. However, interpreting changes in measured resistivity in terms of system dynamics remains challenging due to the complex relationships between thermodynamic and petrophysical parameters and resistivity. Most existing relationships have limited geological applicability and often rely on empirical coefficients derived from laboratory analysis. To contribute to this issue, a new machine learning approach based on a Random Forest algorithm is proposed to predict subsurface resistivity values from numerical simulations of the system dynamics. The aim is to establish a relationship between the petrophysical/thermodynamic variables of the numerical model and the 3D electrical resistivity imaging of the study system obtained from field geophysical surveys. Such a relationship could be used to predict temporal variations in resistivity distribution in response to changes in simulated thermopetrophysical conditions. Comparison between predicted and field resistivity data would ultimately validate the current dynamics of the system, providing a powerful additional tool for resistivity monitoring of natural degassing systems. The application of the proposed approach to two CO2-dominated degassing areas in southern Italy resulted in good resistivity prediction accuracy for both the test datasets, showing a significant improvement in resistivity prediction compared to the use of conventional techniques.