Uncertainty Quantification of the Continuous and Discontinuous Geomechanical Response in Over-exploited Aquifers

Groundwater over-exploitation has resulted in various adverse geomechanical consequences, notably land subsidence and the emergence of earth fissures. Despite extensive efforts to predict these geohazards, many studies have overlooked the uncertainties stemming from aquifer properties and geological conditions. Therefore, this thesis aims to quantify associated uncertainties through the integration of numerical modeling approaches, monitoring techniques, surrogate models, and statistical methods. Prior research has underscored the importance of accurate predictions in land subsidence, necessitating proper characterizations of primary parameters such as hydraulic conductivity, soil compressibility, or specific storage. In the context of homogeneous aquifer systems, a surrogate-model-based Bayesian inversion method is proposed. This method leverages a coupled variably-saturated groundwater flow and geomechanical model alongside interferometric synthetic aperture radar (InSAR) data to infer the posterior distribution of these aquifer parameters. To alleviate computational burden, sparse grid collocation is employed to construct cost-effective surrogate models for efficient Markov Chain Monte Carlo (MCMC) sampling. Application of this framework in the Alto Guadalentin Basin, Spain, successfully characterizes hydraulic conductivity and compressibility. Additionally, this thesis develops a decoupling methodology to characterize heterogeneous parameter distributions in confined aquifer system. In these cases, Bayesian inversion remains computationally challenging. Here, the focus lies on characterizing specific storage, while hydraulic conductivity has been previously deduced by incorporating piezometric data into groundwater flow modeling. Based on groundwater solution and mesh configuration, specific storage at InSAR observational points is derived using a one-dimensional Terzaghi consolidation equation. Subsequently, the Kriging method is employed to estimate the spatial distribution of specific storage by interpolating these computed "observations". The efficacy of this methodology is demonstrated in the Gediz River Basin, Turkey. In addition to land subsidence, earth fissure formation is investigated using a novel numerical approach. By incorporating interface elements (IEs) into the continuum finite element (FE) model, the present thesis replicates the formation and development of multi-fissures in the Guangming village, China. Numerical outcomes reveal the mechanisms behind fissure formation, where tensile stresses cause opening and shear stresses lead to sliding. Furthermore, global sensitivity analyses are conducted using three different methods: Monte Carlo method, polynomial chaos expansion (PCE), and gradient boost tree (GBT). All analyses highlight the significant impact of ridge slope in triggering earth fissures given specific piezometric level declines. Overall, this thesis significantly advances our understanding of the complex behaviors of over-exploited aquifer systems and offers effective methodologies to address associated uncertainties. These contributions are crucial in the development of sustainable groundwater management strategies, especially in face of escalating water demand.