Assessment of heterogeneous hydraulic properties in natural aquifers at the intermediate scale

One of the most relevant issues in groundwater hydrology is the assessment of heterogeneous hydraulic properties of natural aquifers at the intermediate scale. The intermediate scale includes, as a general definition, aquifers characterized by a ratio L/B, L being a linear measure of the horizontal extent, and B the average thickness, between one (local scale) and approximately one thousand (regional scale). In Italy groundwater represents almost eighty percent of drinkable water supply, therefore the control of groundwater pollution as well as an optimal planning of the well production activity are essential steps in groundwater management that are generally developed at the intermediate scale. The assessment of unknown geological properties affecting groundwater dynamics and the quantification of the related uncertainty is thus one of the most relevant issues, known in literature as parameter estimation or inverse modelling. Regional and local scales have been largely studied by the scientific community, although the local scale is typically less exploited than the regional. The intermediate scale, instead, has been the subject of a relatively limited number of publications and is still characterized by relevant open problems, like, for example, the optimal choice of the numerical model to adopt or the definition of a proper parameter estimation approach. Moreover, a limit is often given by the lack of experimental sites available for testing the new research developments. As a result, too often the research is not applied in real-world problems. This thesis is subdivided in four chapters. In Chapter 1 we introduce the multi-scale experimental site of Settolo, an alluvial unconfined aquifer in north-eastern Italy that constitutes, as far as we know, the first multi-scale groundwater test site in our country. The installed instrumentation and the collected data are described, together with a preliminary data interpretation. In Chapter 2 we focus on groundwater modelling at the intermediate-scale. We calibrate a homogeneous 2D groundwater model to reproduce the Settolo field piezometric data and we explore the differences with a 3D model implemented with the same parameters and boundary conditions. We compare again the two models with spatial heterogeneity in the parameters. Because of the very similar performance of the 2D and 3D models we conclude that, in our case, the 2D model is suitable to describe the Settolo aquifer behavior at the intermediate scale; therefore, it can be effectively adopted for inverse modelling approaches and data assimilation. Chapter 3 introduces a novel global optimization algorithm implemented in a geostatistical framework to calibrate groundwater models. The proposed parallel algorithm is tested and validated under various conditions, with both homogeneous and heterogeneous parameter distributions, showing robustness and efficiency in solving the inverse problem, as well as scalability to high performance computing. Finally in Chapter 4 we focus on the geological facies-heterogeneity, not suitable to be described by classic geostatistical approaches like, for example, variogram based models. We approach the inverse problem with a strategy that combines multiple point geostatistics and sequential data assimilation with the ensemble Kalman filter, involving different kind of data, i.e., satellite images, electrical resistivity tomographies and time series of piezometric heads recorded in different locations. This approach is tested for the first time on a real-world case, and its advantages and limits are shown