Impact of seasonal variations and transient colmation layer properties on bacteria and virus transport in bank filtration

Bank filtration is a sustainable source for drinking water production in urbanized regions but is increasingly at risk by contamination with pathogenic bacteria and viruses from surface water receiving wastewater discharge. While recent advances have improved our process understanding for pathogen transport on laboratory scale, simulations and predictions on field scale under transient conditions, as in bank filtration, are still highly uncertain. To improve our understanding on field scale, we performed a sampling survey over 16 months at an observation well transect in a heterogeneous sand-gravel aquifer of an active bank filtration waterworks at the river Rhine in Germany. Water samples were collected from the river, the production well, and 4 multi-level observation wells. Samples were analysed for main anions/cations, and hygienic indicators (E. coli and coliform bacteria via plate counts, coliphages via plaque assay, and adenoviruses via ddPCR). A two-dimensional reactive transport model was set up using PFLOTRAN to simulate the transport of heat and dissolved species, aerobic respiration, denitrification, and colloid-based transport of bacteria and viruses. For the latter, adsorption to and desorption from the sediment, straining, blocking, and inactivation are considered. Model parameters were estimated from prior knowledge of the site or calibrated with the obtained data using particle swarm optimization.

Field observations show a strong seasonal variation of river hydraulics with up to 8 m difference in water level, a prolonged low in the summer/fall and short-termed river level increases in the winter. Aerobic respiration was strongly controlled by the temperature variation (6-24°C in groundwater), leading to an increase in oxygen consumption and limited denitrification during the warm summer/fall. Bacteria and virus concentrations in the groundwater were elevated following a flood in the first winter (up to 500 MPN/100mL coliforms, 2 PFU/100mL coliphages, 1000 copies/100mL adenovirus). Measurable concentrations were still observed during the summer (e.g., up to 10 MPN/100mL coliforms, 0.7 PFU/100mL coliphages, 500 copies/mL adenovirus), but concentrations were below the detection limit for most of the second winter, where no significant flood occurred. In the well closest to the river (40 m distance), the concentration reduction compared to the river varied over time between 1 to ≥4 log-units for coliforms, 1.5 to ≥3 log-units for coliphages, and 0.5 to ≥3 log-units for adenoviruses. The model results suggest the main driving processes for the variation in the bacteria and virus concentrations are (i) the changing groundwater velocity (driven by river level variations and pumping rate), (ii) occurrence of low dissolved oxygen concentrations which lower inactivation, and (iii) transient colmation layer properties (permeability and effective grain size). The colmation layer is affected by reworking of riverbed sediments during floods, bio-clogging during summer, and physical clogging due to constant forced infiltration caused by the bank filtration plant. This is supported by the observation of high bacteria concentrations in the aquifer for a short duration after pumps were reactivated following a 40-day maintenance period. Overall, bacteria and virus attenuation during bank filtration was high, only a strong flood resulted in significantly higher contaminant concentrations in the aquifer.