Geophysical modeling for groundwater and soil contamination risk assessment

The research activity developed for this thesis was focused on development, testing and application of data inversion methods and numerical models for solving environmental problems linked to contaminant detection and transport in soil and groundwater. First, three different approaches for self-potential data inversion, respectively based on spectral, tomographical and global optimization methods, have been proposed and tested on synthetic and field data. As regards the spectral approach, three methods, i.e. classical periodogram method (PM), high-resolution multi taper method (MTM) and maximum entropy method (MEM) have been first applied on SP data, in particular for estimating the depth of the anomaly causative sources. An extended comparative study among the three proposed methods has been performed by applying the spectral methods to SP synthetic data generated by simple geometrical bodies, such as sphere, horizontal and vertical cylinder and inclined sheet. The main results of this numerical analysis is that generally MEM is able to give depth values closer to actual values than those provided by PM and MTM. The effectiveness of the proposed spectral approach has also emerged from the analysis of SP experimental signals. Indeed, a very good correlation of the obtained depth values with those provided by other numerical methods has been found (Rani et al., 2015; Di Maio et al., 2017). Although the spectral approach has proved to be able to provide an accurate depth estimation of the SP anomaly sources, for their full characterization a combination of spectral methods with other inversion methods is required. Hence, an approach based on integration of spectral analysis and 2D tomographic technique has been proposed. In such an approach, MEM has been used to find the source depth and, then, COP distribution has been computed to get information about the polarization angle and the position of the anomaly source. Specifically, from the numerical study performed on synthetic examples of horizontal cylinders and inclined sheets, mathematical equations that link the lines of zeros of COP function with polarization angles and horizontal position of the causative sources along the profile have been found. By using such equations for the analysis of field examples, a good agreement has been retrieved between the estimated source parameters and the results obtained from different numerical methods, which has demonstrated the potentiality of integrating high resolution spectral analysis and tomographic approach (Di Maio et al., 2016a). Finally, in order to fully characterize anomalous bodies responsible of the observed SP signals without any a priori information on their shape, a global optimization approach based on a hybrid genetic-price algorithm (GPA) has been proposed. An extensive numerical analysis on SP signals affected by different percentage of white Gaussian random noise have shown that the GPA is able to provide fast and accurate estimations of the true parameters for all tested examples. In particular, the calculation of the root-mean squared error between the true and inverted SP parameters is found to be crucial for the identification of the source anomaly shape. Finally, applications of the GPA to self-potential field data have demonstrated the effectiveness of such an approach for SP data inversion (Di Maio et al., 2016b). As concerns the development of numerical methods for modeling contaminant propagation in soil and groundwater, two cellular automata have been proposed to simulate diffusion-dispersion processes in saturated and unsaturated conditions. Then, the developed models have been applied to two field cases. The first test area is located in the western part of the island of Crete (Greece) close to the Keritis river and close to an olive oils mills waste (OOMW) deposition pond. In this case, contamination is mainly due to phenol concentrations from OOMW, which can propagate through the soil in vertical direction reaching the saturated zone at a depth of about 5 m or in horizontal direction towards the river. On the basis of the geological and geophysical characterization of the area, a cellular automaton model has been developed for simulating contaminant diffusion in the unsaturated zone. The CA grid represents the vertical section of a three layers model, where each cell describes a small portion of soil characterized by the values of three physical parameters: electrical resistivity, porosity and contaminant concentration. By using a relation rule typical of critically self-organized systems and performing an analysis by varying the diffusion transfer coefficients, the simulation results suggest very likely contaminant percolation in the saturated zone in two specific periods of the year. Finally, a finite element model based on FEFLOW simulation software has been developed for simulating the groundwater flow and contamination transport under unsaturated and saturated conditions of the survey area. The flow calibration has shown a good agreement between observed and computed hydraulic heads, while the study for mass transport and calibration with resistivity and geochemical data is in progress. The second selected test area is located in southern Italy close to the Solofrana river and it is often affected by floods with polluted water and mud. The area has been first characterized by a multidisciplinary study, which has integrated the results of a geoelectrical survey, consisting of resistivity and induced polarization tomographies and SP profiles, with hydrogeological and geochemical data. On the basis of such a characterization, a cellular automaton (CA) model has been developed for simulating contaminant infiltration induced by flooding phenomena. The CA grid represents the vertical section of a three layers model, where each cell describes a small portion of dry or wet soil. By performing an analysis on millions of initial different configurations, the conditional probability distributions of observing the contaminant at the time t and at distance L from the top of the grid have been calculated, and, as a preliminary result, a plot L-t for estimating mean infiltration rates of the contaminant in the three soil layers has been obtained.