Transience of Bank Filtration in Snowmelt-Dominated Basins: from Tracer-Based Characterization to Monitoring and Operational Strategies

Bank filtration (BF) is a pre-treatment technology to produce drinking water that benefits from the natural capacity of the sediments to attenuate contaminants. As BF systems operate in transient conditions, predicting their efficiency is difficult, especially under flood conditions. Meanwhile, little has been done to understand how these systems function in snowmelt-dominated basins, where the hydrological cycle is shaped by springtime freshets and frequent flood events. The objective of this thesis was to reveal the dynamics of groundwater flow patterns at a BF site affected by recurring springtime flood events. The specific objectives were: (1) to distinguish the role between anthropic and meteorological controls on the mixing ratios and on the physico-chemical quality of the raw water, (2) to develop a short time-step monitoring strategy to reveal the spatiotemporal dynamics of the groundwater flow patterns under extreme flood events, and (3) to differentiate the respective role of direct floodwater inputs and temporary subsurface storage of floodwater on the water balance of a lake used as source water for the BF site. First, a time-varying binary mixing model was developed to investigate the competition between anthropic and meteorological forcings on the origin of pumped water at a monthly timescale. Given that the BF sites are operated under highly transient conditions, it appeared crucial to determine the spatiotemporal variability of the origin of raw water due to pumping (anthropic forcing) and water level variations (meteorological forcing). It was shown that pumping schemes can be used to control the origin and quality of the pumped water. The operation of the pumping wells on a continuous basis allowed to outweigh the meteorological control (i.e., water levels), except during flood event. Such pumping scheme at high pumping helped at maintaining the mixing ratios at ~62% of Lake A water and at limiting the total iron (Fe) and total manganese (Mn( concentrations in the pumped water. Conversely, the operation of wells on an intermittent or occasional basis resulted in large variations in mixing ratios (i.e., from 0% to 100%) on the annual timescale and was found to deteriorate the physcio-chemical quality of the pumped water. Second, an optimized combination of environmental tracers, including electrical conductivity (EC), stable isotopes of the water molecule (δ18O- δ2H) and temperature, and a short time-step monitoring, allowed to better understand the daily to monthly dynamics of groundwater flow patterns reorganization during and after an extreme flood event. It is recognized that floods represent a challenge for the operation of BF systems. In fact, the quality of surface water is typically deteriorated, and short travel times are often observed at BF systems during floods. Thereby, BF systems are vulnerable to various contaminations during inundation. The different tracers were found to be complementary and to provide information that can help at anticipating microbial water quality changes and guide the timing and the frequency of more expensive and time-consuming monitoring analyses (e.g., bacteriological indicators) to secure the production of high-quality drinking water. While the evolution of EC and δ18O in the pumped mixed water and at the pumping wells revealed a gradual evolution of the origin of water over a 1-month period, automated measurements of temperature and EC at observation wells allowed to detect changes in the groundwater flow patterns at a daily timescale. Monitoring for environmental tracers additionally allowed to track the impact of a dyke failure on the origin of the bank filtrate. A continuous monitoring of CE and temperature revealed a rapid reorganization of the groundwater flow patterns, leading to a temporary inversion of the hydraulic gradients in the vicinity of the pumping wells and a change in the origin of the pumped water. Lastly, the monitoring of δ18O at the pumping wells helped at depicting the recovery dynamics to normal conditions. As expected, the operation of wells at high rates (i.e., ≥ 1000 m3/day) accelerated the recovery to pre-flood mixing ratios. Third, a volume-dependent stable isotope mass balance was developed to estimate the partitioning between groundwater and floodwater to Lake A (i.e., principal source water to the pumping wells), in order to depict the resilience of the hydrosystem to changes. The model revealed that groundwater inputs dominated the annual water budget, with groundwater representing 60%-72% of the total inputs on the annual timescale. However, when considering the potential temporary subsurface storage (and discharge) of floodwater, the partitioning between groundwater and surface water inputs tends to equalize. This implies that floodwater inputs and storage contribute to increase the resilience of the lake to groundwater quantity and quality changes. Thereby, the recurring floods appear to contribute positively to the hydrosystem and help at securing the sustainability of drinking water production at the BF site. Considering all parts, this thesis provides methodological approaches to enhance knowledge on the functioning of BF systems affected by recurring springtime freshets. This research explored the dynamics of groundwater flows and floodwater storage from the daily to annual timescales at both the BF site scale and hydrosystem scale. Monitoring and operational strategies were applied and tested. They now contribute to the solutions that operators and scientists dispose to manage water quality. This work thus benefits to the water managers and scientists, as it can help at operating existing BF sites and planning future BF projects in snowmelt-dominated basins.