The potential of major ion chemistry to assess groundwater vulnerability of a regional aquifer in southern Quebec (Canada)

Regional hydrogeochemistry of the study area

Groundwater chemistry types were determined from the Piper diagram (Fig. 4). Fifty-six per cent of the samples belong to the (Ca, Mg)–HCO₃, Ca–HCO₃ and Ca–SO₄ water types, mainly present in the Appalachian region where unconfined conditions prevail (see Fig. 3). Thirty-nine per cent of the samples belong to the Na–HCO₃ water type, representing semi-confined conditions in the flat, central part of the study area to captive conditions in the downstream portion. Five per cent of the samples belong to the Na–Cl water type located in the lower portion of the aquifer. The Na–Cl water type represents confined conditions and is associated with a mixing with trapped marine pore-water. Charron (1978), Cloutier et al. (2008, 2010), Beaudry (2013) and Benoit et al. (2014) described similar hydrogeochemical facies for other southern Quebec bedrock aquifers.

Open image in new window

In the recharge area, groundwater chemistry is first controlled by the dissolution of carbonates present in calcareous Quaternary deposits such as tills (Cloutier et al. 2010) and through the first few metres of the fractured bedrock aquifer (Edmunds et al. 2003) which is here dominated by calcareous shaly and silty deposits of Ordovician age (Globensky 1987). In the study area, groundwater typically evolves from (Ca, Mg)–HCO₃ water types with low but nearly equal proportions of Na⁺ and Cl⁻ (representing an initial load of Na–Cl from precipitation) and with equal proportions of Ca²⁺+Mg²⁺ and \({\text{HCO}}_{3}^{ - } + {\text{SO}}_{4}^{2 - }\) (from the initial congruent dissolution of calcareous and dolomitic materials). Slower and deeper circulation, involving long-term water–rock interactions in fractured aquifers, generally involves other processes such as incongruent dissolution of carbonates and cation exchange (Edmunds et al. 1987). For most of the samples taken in the Becancour bedrock aquifer, cation exchange (Ca²⁺→Na⁺) appears to be the controlling process (Fig. 5), leading to the relative enrichment of Na⁺ versus Cl⁻ and relative depletion of Ca²⁺ + Mg²⁺ versus \({\text{HCO}}_{3}^{ - } + {\text{SO}}_{4}^{2 - }\) towards the evolution of groundwater to the Na–HCO₃ type (Cloutier et al. 2010). Na⁺ is released in groundwater by cationic exchange from the clay fraction, whereas Ca²⁺ is depleted from groundwater. For samples with lower (Ca + Mg)/(Na + K) ratios, located in captive groundwater flow conditions, a shift (grey circle Fig. 5) in the total cation content probably indicates a mixing with the seawater end-member. In the study area, Na–Cl is released from marine clays or trapped Champlain Sea seawater (Cloutier et al. 2010), and groundwater evolves to Na–Cl type.

Open image in new window

Nitrate concentrations are relatively low, with a maximum value of 6.1 mg N–NO₃/L (drinking water limit 10 mg N–NO₃/L; MDDEFP 2013) and not related to the well depth (Table 1). These concentrations are similar to those reported in other studies from southern Quebec (MENV 2004). Eleven wells (12.9 %) mainly located in the Appalachian foothills have concentrations exceeding 1 mg/L N–NO₃, i.e. the anthropogenic background (Dubrovsky et al. 2010). Low nitrate concentrations in the study area could be attributed to the high recharge rates (159 mm/year, Larocque et al. 2013) which might induce significant dilution within the aquifer (Andrade and Stigter 2009). They could also be due to the interception of nitrate-rich infiltrated water by agricultural drains (Qi et al. 2011) and to denitrification (see further discussion below).

Hydrogeochemical Vulnerability Index

Considering the hydrogeochemical processes occurring in the study area, plotting the relative differences of Na⁺–Cl⁻ against (Ca²⁺ + Mg²⁺) − (\({\text{HCO}}_{3}^{ - } + {\text{SO}}_{4}^{2 - }\)) illustrates how groundwater composition evolves from a (Ca, Mg)–HCO₃ type in the recharge area to a Na–HCO₃ type downgradient with gradually increasing confinement conditions (R ² = 0.97; Fig. 6). A Hydrogeochemical Vulnerability Index (HVI) from 1 to 10 was attributed to each of the water samples from its orthogonal projection on the linear regression. The HVI identifies the highest vulnerability scores for groundwater samples near the recharge area ((Ca, Mg)–HCO₃, Ca–HCO₃ and Ca–SO₄ types). It gradually decreases with increasing confinement conditions of the bedrock aquifer, thus changing to Na–HCO₃ type and eventually to Na–Cl type. A HVI map was interpolated (inverse distance weighting) using all the sampled wells (Fig. 7). The highest values of HVI are found in the Appalachian Mountains (HVI above 8) and gradually decrease with the regional groundwater flow, as the Quaternary deposits become thicker and/or more impermeable. Approaching the St. Lawrence River, the least vulnerable areas (HVI between 1 and 5) correspond to the location of thick clay deposits which confine the bedrock aquifer. The HVI generally reflects the bedrock aquifer confinement conditions (see Fig. 3), but also locally highlights the discontinuities in confinement that differ from the regional pattern. For example, the HVI shows probable local recharge areas within dominantly confined conditions (HVI above 8), particularly downstream, in the north-eastern part of the basin and in the Appalachian Piedmont.

Open image in new window

Open image in new window

The HVI can be derived similarly for other hydrogeological contexts where calcareous dissolution from recharge and Ca²⁺–Na⁺ cation exchange dominate the water chemistry. This is a common pattern regarding the evolution of groundwater geochemistry composition along the flow paths (Edmunds et al. 2003; Appelo and Postma 2005). The HVI is expected to be particularly suitable in glacial hydrogeomorphological contexts, i.e. with the presence of calcareous materials found in tills, in sedimentary bedrock rich in clay minerals, and under gradual confinement conditions of the fractured aquifer with marine clays. The HVI was tested in other regions of the St. Lawrence Lowlands in southern Quebec, taking advantage of newly available hydrogeochemical datasets from the Nicolet and lower Saint-François watersheds (Larocque et al. 2015) and in the Monteregie region (Carrier et al. 2013). For these regions, the HVI was built using the same ranges of Na–Cl (y-axis) and Ca + Mg–HCO₃–SO₄ (x-axis) than for the current study area. HVI values larger than 9 were found, respectively, for 83 % and 89 % of wells with detectable nitrate (above 0.1 mg N–NO₃/L) (results not illustrated). In other geological and climatic contexts, it might be necessary to modify the axes ranges in Fig. 6 to adjust to site-specific conditions.

The HVI provides integrative information on well vulnerability while being relatively inexpensive to implement and easily computed. Similarly to other types of interpolated maps (e.g. maps of hydrogeological contexts interpolated from local drilling data), the uncertainty related to the interpolated surface is directly linked to data point density. For example, because municipal and private wells are rare near the St. Lawrence River (due to high-salinity groundwater), the vulnerability map in the lower portion of the study area is more uncertain. In the case of local groundwater contamination, the method may show heterogeneities related to anthropogenic sources (e.g. Na–Cl pollution from deicing road salts, surface water infiltration into a poorly maintained well cap) rather than to the natural hydrogeological context. A preliminary check was performed to discard results from samples impacted by deicing road salt.

DRASTIC vulnerability map

The calculated DRASTIC indices range from 33 to 179 (Fig. 8). This scope is comparable with existing DRASTIC maps obtained in similar geomorphological contexts in southern Quebec (Champagne 1990; Murat 2000). In the current study, the DRASTIC index in the Appalachians indicates the simultaneous presence of medium (between 76 and 100) and high (between 126 and 150) aquifer vulnerability areas. Low groundwater levels on topographic ridges counterbalance high recharge rates. Areas where the aquifer vulnerability is very high (higher than 150) are characterized by the presence of granular glaciofluvial deposits where the groundwater depth is shallow. These are found in the valley bottom of the Becancour River in the eastern part of the study area. The index is high (between 126 and 150) to very high (higher than 150) in the central part of study area due to high groundwater levels in a flat topographic context, as well as to the presence of regressive and aeolian sand deposits. In this part of the study area, recharge rates can be high but are counterbalanced by the presence of silty deposits and peatlands (between 76 and 100) which lowers parameters S (soils) and I (impact of the vadose zone). Low aquifer vulnerability areas (between 33 and 75) are located in the lower part of the basin where thick clay deposits are present.

Open image in new window

It is not possible to integrate complex sequences of overlying Quaternary deposits when using the DRASTIC index to estimate the vulnerability of a bedrock aquifer. This is due to the fact that it uses only two parameters for the overlying unconsolidated sediments, i.e. one parameter for soil type (S) and another parameter for the impact of the vadose zone (I). Recharge is usually the main vector for solute transport through the vadose zone, as highlighted in other studies the importance of recharge in aquifer vulnerability estimation (e.g. Rupert 2001; Nobre et al. 2007). This is reflected in the index, where four out of seven parameters relate to infiltration and recharge processes (R, S, T and I). On the DRASTIC map (Fig. 8), the most extended and vulnerable aquifer zones are found in the Appalachian Piedmont, while hydrogeochemistry suggests that recharge areas are predominantly located in the Appalachian Mountains. High aquifer vulnerability in this area is caused by high S and I parameters in the presence of sandy soils. However, these regressive and aeolian sands are often superficial deposits covering thick till deposits considered as aquitards. This underlying till probably offers a good protection for the underlying bedrock aquifer, which is not taken into account by the DRASTIC index.

Comparing HVI and DRASTIC indices

The two vulnerability indices (Figs. 7, 8) do not show the same spatial distribution because they do not integrate the same parameters and processes. The HVI reflects a cumulative sequence of hydrochemical processes, from the recharge area to the aquifer domain where renewal rates are low. It can be considered as an integrated representation of the flow lines from the recharge area to the pumped well and reflects borehole vulnerability. The DRASTIC index is mostly based on parameters responsible for vertical flows from the surface to the aquifer. It maps areas of potential groundwater contamination and does not consider flow line distribution or convergence of diffuse pollution to a well, nor does it take into account groundwater flow dynamics. For instance, an area covered with impermeable sediment will be associated with a low DRASTIC index, but groundwater at this site may be affected by contamination from neighbouring upstream vulnerable areas where recharge occurs.

Nitrate is highly leachable in groundwater and is often chosen to trace anthropogenic impacts on groundwater resource at regional scales (Rodriguez-Galiano et al. 2014). It is also the contaminant that most frequently exceeds drinking water standard (e.g. MENV 2004; US Environmental Protection Agency 2005; Stuart et al. 2007). However, because it is subject to denitrification, nitrate is not a perfect tracer. Previous studies detailed the processes involved for denitrification due to oxic/anoxic conditions (Korom 1992; McMahon et al. 2004) which are associated with microbial (McMahon et al. 2008) or pyrite reduction (Böhlke et al. 2002) in groundwater. In the current study, denitrification rates were not estimated because of the density of sampling locations and because of the relatively low nitrate concentrations measured.

The HVI and DRASTIC indices were compared against measured nitrate concentrations to estimate the predicting capacity of the vulnerability method (note: the HVI computation does not include nitrate). The occurrence of nitrate in groundwater depends (1) on the spatial distribution of non-point sources of contamination at the land surface, (2) on the capacity of nitrogen to be mineralized and migrate vertically through the unsaturated zone towards the aquifer and (3) on groundwater flow paths within the saturated zone. Since the DRASTIC index reflects (2) while the HVI reflects (2) and (3), comparing the HVI and DRASTIC vulnerability scores with the occurrence of nitrate provides a means to evaluate how each of these three mechanisms influences nitrate concentrations at a given well. To compare the two vulnerability indices with the nitrate and within a single range, DRASTIC indices (from 33 to 179) were linearized to values between 1 and 10 (Fig. 9). For samples with detected nitrate, the linearized DRASTIC index ranges between 4.5 and 8.7, but does not show a gradual increase pattern with nitrate concentrations. This result suggests that the DRASTIC index is relatively ineffective for evaluating more and less vulnerable areas in the study area. In sharp contrast, the HVI shows a consistent trend with the occurrence of nitrate within the Becancour area, with 82 % of all wells having detectable nitrate concentrations (above 0.1 mg N–NO₃/L) having a vulnerability score of greater than 9.

Open image in new window

Because both methods estimate the intrinsic vulnerability, neither considers the presence of a contaminant source at the surface and the difference between their capacities to identify the presence of nitrate in groundwater is linked to one of the other two causes. The better nitrate/vulnerability index correspondence for the HVI method is probably explained by the fact that major ions geochemistry reflects groundwater flow dynamics which are not considered in DRASTIC. The inclusion of parameters considered representative of water and contaminant flow through the unsaturated zone (i.e. A and C parameters) is not sufficient for DRASTIC to correlate better with nitrate concentrations. Using concentrations from other anthropogenic contaminants, performing tracer tests or groundwater flow modelling on the studied aquifer would provide a more complete test of the HVI method capacity to identify regional-scale groundwater vulnerability.