|dc.description.abstract||Groundwater models can be useful tools to support decisions regarding the management of public water supply wells. Scientific progress and the availability of increasingly powerful computer resources provide a continuous opportunity for improving the way numerical models are applied for this purpose. In this thesis, numerical groundwater models were applied to address relevant questions regarding the management of water supply wells in two distinct glacial aquifers in southern Ontario, Canada. The objective is to propose science-based methods that can be applied in day-to-day practice in the context of source water protection. Three specific issues were addressed:
(1) Time lag in the unsaturated zone: A simplified method was proposed to assess the importance of the unsaturated zone in delaying the effects of changes at ground surface on water supply wells (i.e., unsaturated zone time lag). This assessment is important because it influences field and modelling efforts to estimate well vulnerability to contamination and impacts due to changes in land use. The proposed method is based on estimations of travel time in the saturated and unsaturated zones, and provides a formal framework for an intuitive approach. For the studied case, the delay in the unsaturated zone was deemed to be significant, representing on average ~ 11 years or ~ 53% of the total travel time from ground surface to receptor. Travel times were estimated using approaches with different levels of sophistication, to evaluate the usefulness of simplified calculations. Such calculations lead to the same overall conclusion as more sophisticated and time-consuming approaches. However, when assuming limited knowledge of soil properties, common at earlier stages of most investigations, these simplified techniques generated inconclusive results.
(2) Uncertainty in capture zone delineation: A simple method was proposed to address the issue of uncertainty in capture zone delineation. This method considers uncertainty at two different scales: local (parametric) and global (conceptual). Local-scale uncertainty is addressed by using backward transport simulation to create capture probability plumes, with probabilities ranging from 0 to 100%. Global-scale uncertainty is addressed by considering more than one possible representation of the groundwater system (i.e., multiple model scenarios). Multiple scenario analysis accounts for more than one possible representation of the groundwater system, and it incorporates types of uncertainty that are not amenable to stochastic treatment (e.g., uncertainty due to conceptual model, to different model codes and boundary condition types). Finally, the precautionary principle is used to combine capture probability plumes generated by different scenarios. As a result, two maps are generated: One for wellhead protection, and another for selection of priority areas for implementation of measures to improve water quality at the supply well. For the studied case, three models with different spatial distributions of recharge but with similar calibrations were considered, exemplifying the issue of non-uniqueness. The two maps obtained by the proposed method were significantly different, indicating that recharge distribution represents a major source of uncertainty in capture zone delineation.
(3) Effects of agricultural Beneficial Management Practices (BMPs) in supply wells: A numerical model framework was used to estimate the effects of measures to reduce nitrate leaching to groundwater from agricultural activities (i.e., Beneficial Management Practices, or BMPs). These measures were implemented in 2003 (~ 10 years ago) at the Thornton well field (Woodstock, Ontario, Canada) to improve well water quality. This case study is based on extensive field work characterization from previous research, and allows the discussion of practical issues related with data collection and interpretation (e.g., different techniques for generating mass loading distributions were compared). Regional flow was simulated in a 3D larger-scale saturated flow model, while variably-saturated flow and transport were simulated in a 3D smaller-scale, more refined grid. A vertical 1D model was used to define the discretization in the unsaturated zone of the variably-saturated flow and transport model. Results indicate that the adoption of BMPs in selected areas can be an effective strategy to improve water quality in supply wells impacted by non-point source contaminants. For the Thornton well field, the currently adopted BMPs are estimated to reduce concentrations from ~ 9.5 to ~ 7.5 mg NO3-N/L. Water quality at the wells are predicted to respond after 5 to 10 years after implementation of BMPs, and are expected to stabilize after 20 to 30 years Management scenarios with further reductions in nitrate concentration are expected to further reduce concentrations by ~ 0.4 to ~ 0.8 mg NO3-N/L. The proposed framework can be adapted to design and evaluate BMPs for similar problems and for other non-point source contaminants.
Some insights were common to all three issues discussed and can be useful to practitioners involved in source water protection studies: (1) Reliable recharge estimations are essential for the management of water supply wells; and (2) The use of multiple models should be encouraged to increase the understanding of different aspects of the system, assess uncertainty and provide independent checks for model predictions.||en