Parametrization and model development for simulating redox reactions related to degradation of natural organic matter in managed aquifer recharge

Managed aquifer recharge (MAR) is commonly used as a treatment method to remove organic matter from surface water for a drinking water supply by filtering the water through soils in an aquifer. Groundwater, solute transport, and natural organic matter (NOM) degradation models are available for describing the processes, but their applications for practical assessment of MAR sites and their design have been limited. The objective of this study was to model the NOM processes, quantify the impacts of individual drivers on degradation, and identify the model parameters mainly controlling concentrations of organic matter, oxygen, Fe, and Mn. The methodological approach was to construct a 1-D reactive transport model for total organic carbon (TOC), dissolved oxygen (DO), Fe, Mn, and heat to address the key NOM processes in a lake-aquifer system, where lake water naturally infiltrates into an aquifer. The model was implemented with generalized likelihood uncertainty estimation (GLUE) and Sobol’s method to test the parameter sensitivity.The results showed that in addition to reactions in the aquifer, the lake bottom sediment model played a key role in bank infiltration through its control of DO concentrations. The model was able to represent the diminishing effect of strong temperature seasonality and smoothening of the temperature-related concentration variations in the aquifer. On the other hand, the model produced seasonal TOC dynamics close to the lake-aquifer interface, which was not detected in measurements. With GLUE, the TOC and DO concentrations in the model were most sensitive to the seepage velocity, the hydrodynamic dispersivity, and the reaction rate parameters controlling the oxygen-related degradation of TOC in the lake sediment and the aquifer. Sobol’s total order indexes also showed sensitivity to parameters that control the temperature dependency of reaction rates. In addition to these parameters, Mn and Fe concentrations were sensitive to reaction rate factors for the dissolution of Mn in the aquifer, but the result was less clear. The resulting identification of the key parameters provides a benchmark for the calibration of bank filtration models in the Nordic context.