Uncertainty Analysis and Identification of Key Parameters Controlling Bacteria Transport Within a Riverbank Filtration Scenario

Managed aquifer recharge through bank filtration is an important method to produce sustainable drinking water. Yet, water quality related to transport of pathogens (bacteria and viruses) into groundwater systems from surface waters can be a matter of concern, especially in urbanized regions. Based on a 1-year monitoring campaign, a reactive transport model was developed for bacteria transport at a riverbank filtration site located in Germany. The model allows simulating advective-dispersive transport and relies on the colloid filtration theory to mimic attachment and detachment of bacteria to and from the sediment in addition to inactivation, straining and blocking of bacteria. Due to the complexity of the investigated processes, the reactive transport model is characterized by a high level of parametrization, encompassing parameters driving flow as well as solute and colloid transport. A global sensitivity analysis has been applied to identify the most relevant model parameters with respect to piezometric pressure heads, groundwater temperature, and concentrations of chloride, oxygen, coliforms, and Escherichia coli. The model has been calibrated within a stochastic framework, to provide model parameter estimates and to quantify their uncertainty. Our results suggest that bacteria transport models are highly sensitive to inactivation coefficients, straining coefficients, and bacteria size. Permeability of the colmation layer at the riverbank is a key factor for bacteria transport through its influence on residence times. Seasonal variability of boundary conditions, especially anoxic aquifer conditions in the summer and high groundwater flow velocities during flooding periods, resulted in a reduction of inactivation and increased bacteria concentrations at observation wells.