Evaluation of uncertainty in kinetically controlled and unsaturated heterogeneous reservoirs under wastewater-groundwater mixing

Effective management of Managed Aquifer Recharge (MAR) systems is strongly influenced by the interplay between geological reservoir features and kinetically controlled geochemical reactions. The former are often approximated by a simple geometrical distribution of hydrofacies, while the latter are approximated as unreactive systems or as systems that always reach geochemical equilibria. Fluid-rock interactions induce modifications of the physical-chemical characteristics of the mineral porous medium, the dissolved gases compositions, and the aqueous solution. This setup induces modifications in preferential flow paths, pollutant residence time, and pollutant persistence in the reservoir. The numerical simulation of such multiphase systems is thus challenging due to the combined nonlinear and time-dependent effects acting on them. The presented results focus on the evaluation of the multifaceted uncertainty through a sensitivity analysis, derived from the quantification of the individual impact of: (i) the adaptive spatial discretization resolution, (ii) the kinetically controlled processes, and (iii) the heterogeneity uncertainty in multiphase reactive transport simulations. A MAR system is modelled over a highly heterogeneous geological section, accounting for reactive processes associated with water-rock-gas interactions. These processes are evaluated through the Transition State Theory. The workflow involves (i) geometric spatial discretization of the reservoir in the form of an unstructured mesh, coupled with geostatistical generation of porosity and permeability fields (as continuous and categorical variables, respectively) using MUSE software (Miola 2025, PhD thesis & EGU25); (ii) segmentation of the domain into homogeneous regions via FSUM (Sorgente et al. 2026, Computers & Geosciences) and localized mesh refinement with an increase of details over hydraulic impedance surfaces; (iii) conversion of stochastic geological models into data formats, and (iv) execution of parallel multiphase reactive transport simulations with PFLOTRAN (Hammond et al. 2014, Water Resources Research). The entire process is managed through EWOPE (Miola et al. 2026, Computers & Geosciences), an open-source computational workflow tracker, which ensures full reproducibility and traceability by recording metadata, enabling the backward reconstruction of computational history and a deep analysis of the single multi-realization computations, the core of the uncertainty evaluation. Simulation results indicate that a MAR can be planned and managed by considering a multi-scenario involving variations in unsaturated medium properties, pollutant arrival times, and variations in the flow patterns, from a probabilistic point of view.