Characterizing Aquifer Properties Through a Sparse Grid-Based Bayesian Framework and Insar Measurements: A Basin-Scale Application to Alto Guadalentín, Spain
Please login to view abstract download link
Aquifer characterization is essential for predicting aquifer responses and ensuring sustainable groundwater management. In this study we develop a sparse-grids-based Bayesian framework to infer the hydraulic conductivity and the soil compressibility of over-exploited aquifer systems using Interferometric Synthetic Aperture Radar (InSAR) ground displacement datasets and piezometric records. The phenomenon is described by a poroelastic model (consisting of a set of coupled, nonlinear PDEs) that describes the interplay between groundwater depletion and soil deformation through the explicit quantification of the porosity change. The Bayesian inversion approach enables to compute the posterior distribution of the uncertain model parameters. However, exploring this posterior using Markov Chain Monte Carlo is computationally prohibitive due to the substantial cost of solving the poroelastic model. To overcome this issue, we use sparse-grids surrogate models to approximate such solutions. The methodology is applied to the Alto Guadalentín basin, Spain, where long-term aquifer exploitation has led to a lowering of the water table larger than 100~m causing impressive land subsidence, with rates up to 15~cm/yr as evidenced by InSAR. The results demonstrate that integrating InSAR data significantly enhances the characterization of the aquifer properties, with the resulting numerical simulations aligning well with available observations.