Speaker
Description
Groundwater and other subsurface resources play a fundamental role in modern water and energy systems, yet their exploitation can induce significant geomechanical responses such as land subsidence. In aquifer systems, overexploitation—defined as extraction rates exceeding natural recharge—can lead to substantial declines in piezometric levels, compaction of aquifer deposits, reduction of porosity and storage capacity, and progressive land subsidence. These processes may compromise soil stability, threaten infrastructures, and reduce hydraulic safety. Physically based geomechanical models provide an effective framework for simulating the coupled processes governing subsurface fluid flow and deformation. However, their predictive reliability is strongly affected by uncertainties in the parameterization of constitutive models describing the hydraulic and mechanical behavior of geological formations. In this work, we present an integrated modeling framework combining physically based poromechanical simulations with uncertainty quantification, sensitivity analysis, and data assimilation techniques to improve the characterization and predictive capability of subsurface systems.
The approach relies on three-dimensional fluid–poromechanical models that simulate the interaction between groundwater flow and subsurface deformation through explicit representation of porosity changes. Aquifer and reservoir properties, including hydraulic conductivity and compressibility, are constrained using multiple observational datasets, such as piezometric measurements and satellite-based surface displacement (InSAR). Parameter estimation and uncertainty reduction are performed within a Bayesian framework, integrating inversion techniques with ensemble-based data assimilation methods. To overcome the high computational cost associated with repeated simulations of nonlinear geomechanical models, surrogate modeling techniques are introduced. Sparse-grid approximations and generalized Polynomial Chaos Expansion are employed to approximate the forward model response, reducing the computational burden of Bayesian inversion and data assimilation. Applications to both a synthetic deep hydrocarbon reservoir and the Alto Guadalentín aquifer system (Spain) demonstrate that integrating satellite deformation data substantially improves the characterization of subsurface properties and enhances the robustness of geomechanical predictions while maintaining computational efficiency.