A geological cross-section through the aquifer system is presented in Fig. 5 (location shown on Fig. 2). Colorado Group underlies the geological units in the study area. It is a 500-m thick regional aquitard and contains several thin sandstone beds, the most significant being the Bow Island Sandstone. The Bow Island Sandstone is about 25 m thick and is separated from the overlying Milk River Formation by approximately 400 m of shale (Swanick 1982; Phillips et al. 1986; Hendry and Schwartz 1988).
The Milk River Formation is about 100 m thick on average and is subdivided into three members. The basal Telegraph Creek Member is a transitional unit between the Colorado Group and the sandstone of the middle member Virgelle. The Virgelle Member constitutes the MRA. The upper member of the Milk River Formation is the low-permeability Deadhorse Coulee Member.
The Milk River Formation not only outcrops or subcrops following continuous and narrow belts on both sides of the Sweet Grass Arch, but also around the Sweet Grass Hills and in southern Alberta near the international border (Fig. 2). The Virgelle Member outcrop is well recognizable west of the Sweet Grass Arch, as it shows many high escarpments (Collier 1930; Zimmerman 1967).
The Milk River Formation is overlain by the Pakowki Formation (Claggett Formation in Montana), which is a 130-m-thick aquitard. The Pakowki/Claggett aquitard thins towards the west and north-west in Alberta (Pétré et al. 2015). The Belly River Group (Judith River Formation in Montana) overlies the Pakowki/Claggett Formation (Fig. 5). A detailed description of the stratigraphy and hydrostratigraphy of these Upper Cretaceous units in the study area is provided in Pétré et al. (2015).
The study area is covered by glacial drift, except on the topographic highs (such as the Cypress Hills and the Sweet Grass Hills) and in the coulees (Williams and Dyer 1930; Colton et al. 1961; Fullerton and Colton 1986; Robertson 1988). As the glacial drift consists mainly of low-permeability till, the surficial deposits in the study area generally do not constitute productive aquifers (Borneuf 1976; Robertson 1988); however, the buried valleys can form very productive aquifers in southern Alberta (Farvolden et al. 1963). Although the content of the buried valley is generally heterogeneous, making their aquifer potential difficult to predict, the fill material of the buried valleys in the study area is predominantly sand and gravel (HCL consultants 2004; Cummings et al. 2012). The Whisky Valley Aquifer (west of the Town of Milk River) lies in the bottom of the Whisky buried valley. This sand and gravel aquifer is connected to the MRA locally (Golder Associates 2004).
Transmissivity and hydraulic conductivity of the aquifer
The transmissivity (T) of an aquifer is generally obtained from aquifer tests. It is a fundamental parameter in the characterization of a groundwater resource as well as an important control on groundwater flow. The first transboundary map of the T (in m2/s) of the MRA was produced by compiling 133 T data points compiled from Meyboom (1960); Persram (Alberta Environmental Protection, Hydrogeology Branch, unpublished report, 1992, cited by AGRA Earth and Environmental 1998); Zimmerman (1967); Tuck (1993); Levings (1981); Norbeck (2006) and Water Right Solutions Inc. (Cool Spring Colony - Application for Beneficial Water Use Hydrogeologic Assessment, submitted for Water Right Permit No. 40G 30045714, 5 pp, unpublished report, 2009). The T data and the spatial distribution of the log T values (interpolated by Empirical Bayesian Kriging in ArcGIS) are shown in Fig. 6. The highest T values are present around the Sweet Grass Hills and in the south-western part of the aquifer in Montana. Transmissivity values range from 1 × 10−4 and 3 × 10−2 m2/s, west of the Sweet Grass Arch (Zimmerman 1967; Norbeck 2006) and from 2 × 10−4 to 4 × 10−3 m2/s in the Sweet Grass Hills area (Tuck 1993), whereas in southern Alberta, T ranges from 1 × 10−6 to 5 × 10−4 m2/s (Meyboom 1960; Persram (Alberta Environmental Protection, Hydrogeology Branch, unpublished report, 1992, cited by AGRA Earth and Environmental 1998). Faulting and fracturing (secondary porosity) around the igneous intrusion of the Sweet Grass Hills have increased the T of the Virgelle Member locally (Tuck 1993). In the south-west corner of the study area, Zimmerman (1967) indicates that the structural deformations may have greatly affected the T of the Virgelle Member, which transmits water mainly through fractured sandstone.
In the south-eastern part of the study area in Montana, east of Chester, the T is low (1 × 10−6 m2/s). In southern Alberta, a high T zone extends from the USA border, around Aden, to Lake Pakowki and then north near Foremost. An area of low T is located in the central part of the study area, between the towns of Foremost and Skiff. West of this low T area, the T values increase over a limited extent but do not attain values as high as in the east. Figure 6 shows a central low-transmissivity area, surrounded by two corridors of higher transmissivity as previously evoked by AGRA Earth and Environmental (1998).
Potentiometric map, groundwater flow paths and artesian conditions
Groundwater flow directions in the MRA, recharge and discharge areas can be inferred from the potentiometric surface of the aquifer. Parts of the potentiometric surface of the MRA were mapped by Meyboom (1960); Borneuf (1974); Toth and Corbet (1986) and AGRA Earth and Environmental (1998) in southern Alberta and by Zimmerman (1967); Levings (1982a), and Tuck (1993) in northern Montana.
Based on these historical potentiometric maps, groundwater flow directions were shown on two-dimensional (2D) vertical cross-sections by Pétré et al. (2015). Previous potentiometric maps are all limited by the international border, preventing a complete representation of potentiometric conditions in the aquifer, especially close to the international border.
The first transboundary potentiometric map of the MRA (Fig. 7) was compiled from historical maps from Zimmerman (1967), AGRA Earth and Environmental (1998), Tuck (1993) and Levings (1982a). Four pressure versus depth profiles provided by Berkenpas (1991) were converted to equivalent hydraulic heads and were also used to complete the potentiometric map at the northern and eastern limits of the MRA in Alberta. Efforts were made to harmonize the various datasets, especially at the USA–Canada border, to properly represent the transboundary groundwater flow characteristics of the aquifer. A dataset of 40 recent water level measurements (2006–2014) collected during the MiRTAP fieldwork (2012–2013), or obtained from public databases—GWIC (2015); Alberta Environment and Parks (2015)—and the Agriculture and Agri-Food Canada, Prairie Farm Rehabilitation Administration monitoring project (PFRA, unpublished report, 2014) was used to validate the transboundary map and confirm the inferred regional groundwater flow patterns.
The highest piezometric heads were measured in the Sweet Grass Hills and in an area north of Cut Bank. These correspond to the recharge areas where the MRA outcrops or subcrops. High piezometric heads were also measured in the area east of Manyberries and in the south-east corner of the study area, indicating a component of groundwater flow laterally from other geological units. This water originates in the Cypress Hills (Toth and Corbet 1986) and the Bears Paw Mountains (Levings 1982a). Piezometric lows are located in the Pakowki Lake area and in the northern part of the study area in Alberta and along Cut Bank Creek and Big Sandy Creek in Montana. Data suggest that these piezometric lows correlate with the talweg of the buried valleys in the study area. The buried valleys may influence outflow of groundwater out of the MRA, as suggested by Borneuf (1976), in southern Alberta. Groundwater flow also converges locally in heavily pumped areas, including the village of Foremost and the Starbrite Colony (15 km south of Foremost, Alberta). The pressure-depth graphs from Berkenpas (1991) indicate the presence of a region with very low groundwater flow (hydrostatic conditions) south-east of Lake Pakowki (along isoline 910 m on Fig. 7).
The MRA is a confined aquifer radially dipping from the outcrop areas which presents flowing artesian conditions in some places. Nearly all the wells in southern Alberta were flowing in the pre-exploitation system (Borneuf 1976; Phillips et al. 1986; Hendry et al. 1991). Currently, flowing artesian areas are located in the northern part of the study area and near Lake Pakowki. These locations are still consistent with the flowing artesian limit first observed by Dowling (1917). Depression of the potentiometric surface in the vicinity of the Etzikom or Chin Coulees are attributed to a number of free-flowing wells that lowered the static water levels (Meyboom 1960; Tokarsky 1974). Recently, four new pressure measurements were obtained on flowing artesian wells in Alberta in 2013 (Fig. 7). These pressures range from 50 to 221 kPa (5–22 m of water) and provide an indication of present-day artesian conditions that still prevail in the MRA.
Although the transboundary potentiometric map of the aquifer is based on four maps representative of various dates and scales, 40 recent data points are consistent with the previous contours maps (except in the areas of heavy pumping where an update of the contours was required). Besides, the hydraulic heads derived from the regional pressure gradients in the northern and eastern limit of the aquifer in Alberta (Berkenpas 1991) are also consistent with the observed hydraulic heads. The potentiometric map is consequently considered as representative at the regional scale.
Transboundary groundwater fluxes
Groundwater flow diverges from the Sweet Grass Hills to the north, east and southeast, whereas west of the Sweet Grass Arch, groundwater flows south-west and north from a groundwater divide located north of Cut Bank (Fig. 7). Based on these observations, two updated transboundary groundwater flow paths were defined: (1) an eastern flow path from the Sweet Grass Hills to the north and (2) a western flow path from the northern part of Cut Bank to the north.
Estimates of transboundary groundwater fluxes were made using the potentiometric and T maps in Darcy-based calculation. The fluxes (Q
east and Q
west for the eastern and western fluxes, respectively) were calculated using Darcy’s law expressed as: Q = T
mean × i
mean × L where Q is groundwater flow (m3/y), T
mean is the arithmetic mean transmissivity near the international border (m2/s), i
mean is the mean horizontal hydraulic gradient (m/m), and L is the length of cross-sectional area through which groundwater flows (m).
The value of the total flux (9.0 × 106 m3/y) in Table 1 should be considered as a maximum. The total flux is probably overestimated because the high transmissivity values in northern Montana correspond to the most productive areas characterized by fault and fractures.
Recharge to the MRA is dominated by infiltration of precipitation in the outcrop and subcrop areas. As most of the study area is covered by glacial drift, recharge waters first enter the unsaturated tills before reaching the Virgelle Member. Robertson (1988) observed that where the low-permeability Deadhorse Coulee Member overlies the Virgelle Member, recharge to the MRA lags and a perched water table can be present.
Limited quantitative information exists concerning the recharge rate of the MRA. An estimate of the effective recharge rate was obtained by assuming the groundwater flux observed at the international border (about 9.0 × 106 m3/y) is solely due to the portion of potential recharge that actually reaches the aquifer. By dividing this total flux by the area of the MRA outcrop/subcrop that contributes to the transboundary flux (9.28 × 108 m2), an effective recharge rate of 9.6 mm/y was calculated. This value is lower than the potential recharge of about 50 mm/y obtained by using the soil-moisture-balance method (Rushton et al. 2006).
Knowing the mean annual precipitation in the outcrop/subcrop area near the border is about 400 mm/y (Climate Canada 2015), the effective recharge represents only 2.4 % of total precipitation. In the prairies of North America, recharge rates range between 2 and 9 % of annual precipitation (Rehm et al. 1982). Using a numerical model, Robertson (1988) obtained smaller values of recharge at the local scale in the subcrop area of < 1 % of total annual precipitation, which were explained by the strong evapotranspiration caused by Chinook (warm dry strong winds). The percentage obtained in the current study is thus close to that of Robertson (1988) and in the lower range of Rehm et al. (1982)
Groundwater inflow in the MRA also occurs through subsurface vertical inflow from other geological units in the topographic highs of the study area, as shown in the potentiometric map. As the aquifer is deep (>400 m) in these areas, this type of inflow would occur at a large time scale and would be less immediate than recharge from precipitation. The flow rates related to these inflows are difficult to estimate and could be better quantified with the use of a numerical model.
In southern Alberta, the natural discharge of the MRA has been identified to occur in springs and seeps located on the southern bank of the Milk River and its tributaries (e.g., Verdigris Coulee, Red Creek) whereas in northern Montana discharge occurs in the Sweet Grass Hills area, along Cut Bank Creek and the Virgelle escarpment (Tuck 1993; Meyboom 1960; Zimmerman 1967; Milk River Watershed Council Canada 2008). Indeed, as shown in Fig. 7, the slope of the potentiometric surface is steeper on the southern side when compared to the northern side of the Milk River. This strongly suggests that the Milk River intercepts a substantial volume of the northerly flowing groundwater in the MRA. To assess the magnitude of this water loss, the mean groundwater flux down-gradient from the Milk River was estimated to about 0.4 × 106 m3/y (using the same calculation used to estimate the transboundary flux) with a mean T of 5.1 × 10−5 m2/s and a mean hydraulic gradient of 0.2 %. Comparing this flux value to the mean groundwater flux up-gradient of the Milk River (9.0 × 106 m3/y) suggests that 96 % of the incoming groundwater flux is intercepted by the Milk River and its tributaries after it crosses the international border. The intercepting role of the Milk River is consistent with Berkenpas (1991). In the south-eastern part of the study area near the border (region A or “no-flow area”, see Berkenpas 1991), the estimate of the elevation of the water yields a similar elevation as that of the Milk River. This suggests that the greater heads up-gradient of the Milk River are not transmitted beyond the river and that the river intercepts the groundwater flow. In Montana, the natural discharge from springs and seeps in the Sweet Grass Hills area was previously estimated at about 0.6 × 106 m3/y (Tuck 1993). These springs are probably rejected recharge which occurs along the island mountain ranges throughout Montana and are not considered part of the regional groundwater flow system. Rejected recharge occurs at the toe of the mountain recharge areas as a result of dramatically reduced transmissivity of the aquifer away from recharge areas (Huntoon 1985).
As no other natural surface outlets have been identified, vertical leakage (or cross-formational flow) from the aquifer through the overlying and underlying geological units must be considered the dominant cause of natural discharge. To estimate the driving forces in place for cross-formational groundwater flow, the vertical hydraulic gradients between the MRA and the overlying and underlying units were calculated and represented spatially in southern Alberta where adequate data were available (Fig. 8). In Montana, cross-formational flow is not considered a major issue near the regional recharge area where infiltration enters the MRA but could be in the south-eastern part of the MRA where there is a lack of data to allow an estimation of the magnitude of this process. Hydraulic heads in the MRA were obtained from AGRA Earth and Environmental (1998), the surficial aquifer heads are from Swanick (1982) and the Bow Island Sandstone heads are from Lies and Letourneau (1995). The average thickness of the Colorado Group between the MRA and the Bow Island Sandstone was assumed to be 400 m based on data from Swanick (1982) and Phillips et al. (1986), whereas the thickness distribution between the MRA and base of the surficial sediments was obtained from the 3D geological model of Pétré et al. (2015). In these calculations, a negative vertical gradient indicates a potential for upward flow, whereas a positive gradient indicates a potential for downward flow. Plots of calculated vertical gradients are presented in Fig. 8. The vertical hydraulic gradients between surficial sediments and the MRA are negative in the vicinity of the Medicine Hat, Skiff and Whisky buried valleys, which indicates an upward flow component from the MRA towards the surficial sediments (discharge conditions). The upward flow component is also confirmed by the presence of many flowing artesian wells in the areas with more negative vertical gradients (Fig. 8a). In the remaining areas, including the central part of the study area and the topographic highs, the vertical hydraulic gradients are positive, indicating a downward flow component from surficial sediments to the MRA. In the central part of the study area, this downward flow is attributed to depressed water levels in the MRA resulting from high pumping rates for long periods, whereas near the USA border it is related to recharge conditions. As depicted in the schematic diagram of Fig. 8c, the high topographic area of the Cypress Hills is also presumed to be a regional recharge area for surficial sediments whose large hydraulic heads induce downward flow through the aquitard overlying the MRA, thus explaining the high potentiometric heads observed east of Manyberries (Fig. 7). This cross-formational flow process was also inferred by Toth and Corbet (1986).
By comparing the hydraulic heads in the surficial aquifer, the MRA and the Bow Island sandstone on a cross-section, Phillips et al. (1986) concluded that the water leaves the aquifer via vertical leakage. From Fig. 8a, it appears that cross-formational flow focused on the buried valley talwegs that have eroded the upper bedrock (Belly River/Judith River and Pakowki/Claggett formations), thus reducing the vertical distance between the MRA and permeable surficial sediments. Besides, the Pakowki/Claggett Aquitard overlying the MRA thins from about 130–100 m in the west and central parts of the study area to 40–60 m west of Skiff. This could also facilitate vertical leakage.
Figure 8a supports the interpretation of Toth and Corbet (1986) who state that the topographic highs in the study areas constitute areas of recharge to the MRA. Toth and Corbet (1986) had also drawn a potentiometric map of the MRA that took into account the presence of buried valleys that were inferred to be zones of flow convergence in the MRA due to the preferential cross-formational upward flow from the MRA where buried valleys are located.
The vertical hydraulic gradient between the MRA and the Bow Island sandstone is positive in southern Alberta (Fig. 8b). This indicates a downward flow component from the aquifer to the Colorado Group throughout southern Alberta; however, the vertical gradient is very small, approximately zero in the Foremost area and is attributed to a depression in the hydraulic head from extensive historical pumping. This downward flow is consistent with the suggestion from Toth and Corbet (1986) and Hendry and Schwartz (1988) that the Colorado Group is a potential sink for water from the MRA. This mechanism could result from the elastic rebound of the shales in the Colorado Group after the 700-m erosion of the land surface during the Pliocene and Pleistocene (Toth and Corbet 1986). Figure 8c summarizes the implications of the mapped vertical hydraulic gradients on cross-formational flow to and from the MRA using a schematic cross-section of the aquifer system.
The same calculations were not applied in Montana due to a lack of data; however, east of the Sweet Grass Arch, in the vicinity of Big Sandy Creek, the heads in the Judith River Formation (equivalent of Belly River Formation; Levings 1982b) are lower than those in the Milk River Formation (Eagle Formation; Levings 1982a). These data indicate an upward cross-formational flow component from the aquifer to the surficial sediments in this area. Although the fluxes related to this upward flow could not be quantified, they are inferred to be small because the MRA is deep in this part of the system.