Identification des sites de prélèvement d'eau souterraine en situation de filtration sur berge par un cadre géochimique et isotopique

Riverbank filtration (RBF) is a method of pre-purification of surface water. It consists of installing a groundwater withdrawal system near-surface water, in a permeable geological formation. The water captured by the pumping well consists of a mixture of groundwater and recently infiltrated surface water. RBF systems have been used for several decades in sites where drinking water requirements exceed the potential yield of aquifers, and where the load of contaminants in surface water is considered too high for direct removal. The passage of water through the banks has the advantage of mitigating the contaminants initially present in the surface water, while allowing the withdrawal of a large volume of water as compared to groundwater sources that do not benefit from such infiltration from nearby rivers or lakes. On the other hand, the reduction in contaminants is particularly sensitive to the rate of mixing between surface water and groundwater and the residence time of surface water in the banks. Any changes in hydrodynamic conditions, whether an increase in surface water level or pumping flow at the well, will have an effect on these parameters. It is therefore important to identify the sites that are in RBF situation, to adequately characterize their vulnerability. The challenge is that a lot of groundwater withdrawal sites are in a situation of RBF unintentionally, as they abstract infiltrated surface water without taking into account the specific vulnerability related to these systems. The main objective of this project is to develop an approach based on environmental tracers, naturally present in the water, to identify on a regional scale the sites that are in the situation of RBF. The electrical conductivity (EC) and stable isotopes of the water molecule (δ2H) were chosen as tracers for their stability (long shelf life, storage conditions at room temperature), low analytical cost and for the simplicity of sampling protocols. The results of this study will not only provide useful information for decision-makers in Quebec, but also develop a methodology that can be exported to other international contexts. To do this, 40 municipal sampling sites were sampled on a weekly to monthly basis for 18 months. The 40 sites were chosen to represent the widest possible range of hydrogeomorphological characteristics (distance to surface water, depth of the strainer, type of aquifer, etc.). As part of this pilot project, facility operators were mobilized to collect the samples themselves, which were then sent for laboratory analysis for δ2H and for the EC. The hypothesis was that these tracers exhibit seasonal variation in surface water, while being stable in groundwater. A time-varying signal, found in the water pumped from a sampling facility, would therefore mean that the surface water signal has propagated through the banks to the well. To properly characterize the seasonal variation in surface water, four rivers with distinct water regimes were sampled at the same time step as the sampling facilities. The 40 sites were classified according to the seasonal variability of the signal in the well water. The time series for 19 sampling facilities showed no seasonal variation. The sites were therefore classified in Group 1: Non-RBF and were considered pumping only groundwater. For 5 sampling facilities, the time series showed a strong seasonal variation, detected at any time of the year. The sites are considered to pump a high proportion of infiltrated surface water and have been classified in Group 2: Permanent-RBF. For 4 sites, the signal varied at only certain times in the year, meaning that the surface water supply was only visible to the pumped water occasionally. These sites have been classified in Group 3: Occasional-RBF, and considered pumping infiltrated surface water only when certain hydrodynamic conditions are met. To quickly represent the sites in RBF situation, the standard deviation on the time series of the δ2H and the EC was calculated. A standard deviation higher than the analytical uncertainty of the tracers is a potential sign of a variation due to the propagation of the surface water signal and not to analytical noise. The standard deviation limits, set at 25 μS/cm for the EC and 1 ‰ for the δ2H, make it easy to identify on an abacus which groups (Non-RBF, Permanent-RBF or Occasional-RBF) each site belongs to. To correctly identify the times in the year when surface water infiltrates to the pumping well, sampling should be carried out at least once a month for 12 consecutive months. As a result, time series data were incomplete for 12 facilities. The vulnerability to microbial contamination of the nine sites classified as RBF was then discussed by comparing episodes of Escherichia coli contamination to the raw water of the sites. Of these sites, 6 showed no sign of E. coli contamination for the duration of the study. While a hydraulic connection between surface water and the well is synonymous with RBF, it is not automatically synonymous with microbiological contamination. Finally, a simplified hydrogeomorphological characterization of all 40 sites was carried out using hydrogeological reports provided by the partners as well as data from the Ministère de l'Environnent et de la Lutte contre les changements climatiques (MELCC). This characterization allowed an initial identification of the links between certain criteria (distance to surface water, depth of the strainer and type of aquifer) and the classification of sites into three groups.