Mapping the Invisible: Geostatistical Breakthroughs from Rainfall Analysis to Aquifer Characterization

This thesis presents innovative methodologies to support groundwater management, focusing on geostatistical methods and inverse modeling procedures. It addresses three key research areas: gap-filling approaches for daily precipitation time series, inversion modeling techniques for subsurface applications, and experimental investigations in controlled laboratory environments. The research begins by introducing a first-time application of geostatistical-based daily precipitation gap filling, tested against a well-established linear interpolation approach. The comparison was demonstrated in the province of Mantova (Italy) and in southern Portugal, but can be easily extended to any area of interest. By introducing monthly semi-variograms instead of daily ones, computational efficiency significantly improves without compromising the accuracy of precipitation estimates. The study also presents the strengths and weaknesses of the application of each procedure. The work then presents three key distinct applications of inverse modeling techniques for subsurface characterization. These applications are demonstrated in synthetic cases through the integration of hydrological and geophysical data. By applying methods such as the Ensemble Smoother with Multiple Data Assimilation, the research demonstrates the power of inverse modeling in estimating complex subsurface properties, such as hydraulic conductivity and electrical resistivity. The method was evaluated (first application) and compared with another geostatistical technique (second application). The traditional geophysical forward model component of the Ensemble Smoother with Multiple Data Assimilation was replaced by a Convolutional Neural Network to optimize processing time (third application). This innovation represents the chapter’s key contribution. The studies reveal that coupling hydrological data with geophysical measurements improves the accuracy and reliability of subsurface models. This combined approach significantly reduces the uncertainty in the estimates, providing a powerful tool for more accurate and detailed aquifer characterization. Finally, the thesis focuses on experimental sandbox investigations. In these experiments, electrical resistivity tomography data and concentration measurements were acquired to estimate subsurface properties in a controlled laboratory environment. The results from the sandbox experiments confirm the findings of the synthetic studies. This research constitutes a significant advance in the integration of geophysical and hydrological methodologies to improve the characterization of subsurface properties, marking the first application of this procedure to laboratory data. Furthermore, it establishes a robust foundation for future experimental studies and suggests several potential improvements in experimental design, data acquisition, and integration of diverse datasets for more accurate subsurface characterization. In conclusion, the work makes significant methodological and practical contributions to support groundwater characterization. It demonstrates the effectiveness of advanced geostatistical, inverse modeling, and laboratory techniques to address key challenges in subsurface and hydrological modeling. The findings suggest several promising directions for future research.