Hydraulic and isotopic data collected from aquifers are routinely used to characterize hydrogeological conditions within sedimentary basins, but similar data from confining units are generally not collected despite their ability to provide insights on important water/solute transport controls. We characterized paleogroundwater flow and solute transport mechanisms across 384 m of Cretaceous shale (aquitard) in the Williston Basin, Canada, using high-resolution depth profiles of water isotopes (δ18O, δ2H). Water samples were also collected from wells installed in the underlying regional sandy aquifer (Mannville Fm; 93 m thick) and from seepage inflows into potash mine shafts (to 825 m below ground). The 1-D numerical transport modeling of δ18O profiles provided insight into large-scale/long-term solute transport in both Cretaceous sediments and the basin. Despite the potential for significant advective migration during glaciations, molecular diffusion appears to be the dominant solute transport mechanism through the aquitard. Simulations suggest average vertical groundwater velocities of <0.05 m/10 ka and an average excess hydraulic head of <10 m; these values are much less than anticipated by successive glaciations. The dominant paleoevent reflected in present-day profiles is introduction during the Pleistocene of glaciogenic meteoric water to the aquifer underlying the shale, likely along an aquifer outcrop area east of the site or through local vertical conduits. Simulations suggest these recharge events occurred during one or more glacial periods. The isotopic profile over the upper 25 m of Pleistocene till and shale is consistent with glacial deposition and transport processes within these units over the Holocene (past 10 ka).
 Knowledge of the hydrogeology and paleohydrogeology of sedimentary basins is critical to understand sources and distribution of natural resources and in the case of groundwater, evaluate the role of these units as protective barriers against contaminant transport. For the most part, groundwater flow and water chemistry within sedimentary basins have been defined using hydraulic, chemical, and isotopic data collected from water-bearing units. However, these water-bearing units represent a small volume of the sediments in the basins and are often interspersed within much thicker, low-permeability clays or shale units (aquitards). Few studies of flow and solute transport within these confining layers (aquitards) have been undertaken.
 Our understanding of the hydrogeology of aquitards is, by necessity, guided by interpretation and numerical modeling of vertical profiles of naturally occurring geochemical and isotopic tracers of pore water (e.g., δ18O, δ2H, Cl−, Br−, and 4He) owing to the fact that hydraulic studies at the field scale are difficult to conduct. These geochemical tracers help define and constrain long-term transport mechanisms. They also preserve a historical record of the impact of major paleohydrogeological changes that occurred in adjacent aquifers. Most geochemical tracer profiles in aquitards have been conducted on near-surface Quaternary deposits and the underlying shales and claystones [e.g., Desaulniers et al., 1981; Hendry and Wassenaar, 1999; Hendry et al., 2000, 2003, Hendry and Wassenaar, 2005; Hendry et al., 2011a]. Few tracer studies have been conducted on indurated and deeply buried argillaceous formations, and those undertaken were to assess their potential as repositories for the disposal of radioactive wastes [see Mazurek et al., 2011, and references therein]. Simulations of tracer profiles in these aquitards generally show that distinctive geochemical and isotopic profiles develop over vertical distances of several tens of meters to hundreds of meters and over time scales of several thousand years to millions of years, with molecular diffusion being the dominant transport process. Further, these profiles provide an archive of the time since the onset of changes in the chemistry of adjacent aquifers. Overall, the application of multiple geochemical and isotopic tracer profiles in Quaternary aquitards and argillaceous formations reveal that natural stable isotopic tracers of water (δ18O and δ2H) and conservative solutes (e.g., Cl−) are most useful for characterization.
 Several recent studies have reported the occurrence of freshwater in laterally extensive aquifers at sedimentary basin margins [Grasby and Chen, 2005; Grasby et al., 2000; McIntosh and Walter, 2005; Person et al., 2003]. The emplacement of this water is largely believed to result from the injection of meltwater into the aquifer outcrop under large hydraulic gradients during periods of glaciation. Numerical simulations of this process have been conducted considering, for example, density effects, mechanical loading, lithospheric flexure, and permafrost formation [Bense and Person, 2008; Lemieux et al., 2008; McIntosh et al., 2011]. The 2-D modeling studies [Person et al., 2007, 2012; Bense and Person, 2008] suggest as much as 500 m of vertical depth penetration occurred in one glacial cycle along the margins of both the Michigan Basin and the Williston Basin (WB) of North Dakota, USA. Dilute waters potentially of glacial origin under confining layers at lateral distances greater than 100 km from an outcrop source, however, have been observed along continental margins [Person et al., 2003, 2007] and at the eastern margin of the Williston Basin into areas currently characterized by high salinity waters [Grasby and Betcher, 2000; Wittrup and Kyser, 1990]. Similarly, the penetration of subglacial recharge has been observed to depths of 500 m or more in England, the Williston Basin, and the Illinois Basin [McIntosh et al., 2012] in stratigraphic sequences.
 The goal of this work was to improve our knowledge of the hydrogeology and paleohydrogeology of a thick, regionally extensive shale aquitard in a sedimentary basin using natural tracer profiles. The specific objectives of this study were to (1) obtain and interpret high-resolution, vertical profiles of the natural tracers δ18O and δ2H through a thick, laterally extensive, argillaceous aquitard, and underlying aquifer in a sedimentary basin; (2) apply 1-D transport modeling of the δ18O profile to gain insights into the controlling solute transport mechanism(s) in the aquitard; and (3) provide constraints on the timing of geologic and hydrogeologic events responsible for tracer profiles in the aquitard and underlying aquifer.
 This study was conducted in the WB, which is a subbasin within the Western Canadian Sedimentary Basin (Figure 1). The WB is a near-circular, intracratonic basin that underlies an area of 250,000 km2 in North Dakota, Montana, and South Dakota in the United States and Manitoba and Saskatchewan in Canada. It contains a near-continuous sedimentary record from the Middle Cambrian to the Cretaceous. The thickness of the sedimentary strata at the center of the WB is about 5000 m. The hydrogeology of the WB has been described using hydraulic head and dissolved chemistry data collected from the aquifers and other water-bearing formations by the oil industry [cf. Bachu and Hitchon, 1996]. However, as noted, the hydrogeology of the numerous thick aquitard units in the WB is poorly understood. One such representative aquitard in the WB is a thick (up to 900 m), regionally extensive, Cretaceous age silt and clay aquitard in the southeast portion of Saskatchewan (commonly termed “shales”).
 The current study was conducted on the Cretaceous shales of the WB at the Mosaic Company's K2 potash mine (5617477.62°N, 295357.89°E), 15 km E of Esterhazy, Saskatchewan, Canada (Figure 1). The data for the study consist of high-resolution profiles of δ18O and δ2H on continuously cored and squeezed samples through 400 m of shale. These data were augmented with δ18O and δ2H measurements from pore water samples obtained from monitoring wells in the underlying Cretaceous Mannville aquifer (392–485 m below ground, BG) and on mine-shaft inflows collected from the Mannville aquifer and deeper formations (451–825 m BG) at the K2 and K1 (located about 15 km NW of K2) potash mines. Because these shales are laterally extensive, and similar shales are present in other basins across North America from the Gulf of Mexico to the Arctic Ocean, our findings are of widespread importance. For example, these shales are a potential reservoir for oil and gas and are perceived as a basal boundary to groundwater flow in the Athabasca oil sands region of Alberta, Canada.
2. Regional Geology and Evolution of the Williston Basin
 Current understanding of the geological evolution of the WB is summarized by Bachu , Bachu and Hitchon , Grasby and Chen , Hacquebard , Nurkowski , Porter et al. , and Price . The WB was initiated during late Proterozoic rifting of the North American craton, developing a succession of marine carbonates and evaporites during the Middle Cambrian to Middle Jurassic. By the Late Jurassic, accretion of allochthonous terranes led to the development of a clastic, shale-dominated, foreland basin succession. During the later Paleocene to early Eocene, the Canadian Rockies foreland fold and thrust belt formed during the Laramide Orogeny about 65 Ma. Subsequent erosion of up to 3 km of Tertiary and Upper Cretaceous sediments during the Tertiary [Beaumont et al., 1993; Hacquebard, 1977] (about 56–2 Ma) produced the modern topographic profile, which was subsequently glaciated during the Quaternary.
 The present-day hydrogeology of the WB is characterized as a gravity-driven SW- to NE-directed groundwater flow system [Bachu and Hitchon, 1996; Hannon, 1987], with groundwater recharge in the Little Rocky Mountains and the Black Hills [Downey et al., 1987; Plummer et al., 1990] and groundwater discharge along outcrops in west-central Manitoba and east-central Saskatchewan [Grasby and Betcher, 2002]. During the Quaternary (about 2 Ma–12 ka), continental Laurentide ice sheets covered the basin with as much as 3 km of ice during multiple glacial stages [Christiansen, 1992; Clark et al., 1996; Ehlers and Gibbard, 2007; Fung et al., 1999; Peltier, 1994]. From stable isotope analyses of water proximal to the aquifer outcrop regions in the basin, Grasby and Betcher  and Grasby and Chen  conclude that this discharging water was meteoric and introduced during a colder climatic period.
3. Materials and Methods
3.1. Hydrogeological Setting
 Two boreholes were drilled near the K2 potash mine. The first (K2A) was approximately 1 km NW of the K2 mine site (total depth of 325 m BG), and the second (K2B) was located approximately 10 km S of the K2 mine site (total depth of 411 BG). The hydrostratigraphy of the Cretaceous formations was the same at K2A and K2B and is summarized in Figure 2. The Cretaceous shale occurred from 11 to 392 m BG and was dominated by shales. It was underlain by the sandy and shaley Mannville Fm (aquifer) from 392 to 485 m BG. At K2A, the shale was overlain by clay-rich Quaternary-aged glacial till (<11 m BG). The Cretaceous shale consisted of the Pierre Fm (11–184 m BG), the First (184–256 m BG) and Second (256–281 m BG) White Speckled shales, the Belle Fourche (281–336 m BG), and the Joli Fou (336–395 m BG).
 The stratigraphy from ground surface to the Prairie Evaporite Fm at the K2 mine is presented in Figure 3. The stratigraphy of the Cretaceous at the K2 mine was the same as that at K2A and K2B, as discussed earlier. The succession of stratigraphic units above the Prairie Evaporite (1076–946 m BG) at the mine consists of the Dawson Bay (946–895 m BG), Souris River (895–781 m BG), Duperow (781–609 m BG), Nisku (609–582 m BG), Three Forks (582–534 m BG), Bakken (534–522 m BG), and Lodgepole (522–485 m BG). The Lodgepole is overlain by the Mannville aquifer (at 485 m BG). The majority of the stratigraphic sequence between the Dawson Bay and the Mannville is dominated by carbonate rock, with interspersed shale sequences.
 Standpipe piezometers (n = 8) completed at shallow depths in the till within 1 km of K2A indicate that the water table fluctuates between >1 and 3 m BG (data not presented). Current vertical hydraulic gradients through the Cretaceous shale, calculated from heads measured using vibrating wire pressure transducers installed at various depths through the till and shale at K2A and manual measurements made in monitoring wells in the Mannville (described later), are downward at 0.22 through the Pierre Shale and 0.03 through the First and Second Speckled Shales, with an overall downward gradient of 0.63 through the 325 m of Cretaceous shale [Smith et al., 2013]. The large magnitude of the observed vertical gradient is believed to be the result of current mining activities. The vertical hydraulic conductivity (K) of the shale measured on core samples collected above the Belle Fourche Formation ranged from (2 to 10) × 10−12 m/s [Smith et al., 2013]. Note that the bulk K of the shale may exceed these values, as has been widely observed elsewhere [van der Kamp, 2001].
3.2. Core Sampling and Analyses
 Continuous core samples were collected between 3 and 321 m BG at K2A in October 2009 using a Failing 1250TD rotary drilling rig and a 3.04 m × 75 mm core barrel. Drill refusal on a concretion layer at 321 m BG prevented cores from being collected to the Mannville. To obtain core samples from the lower 70 m of Cretaceous shale and the top of the Mannville, K2B was drilled in May 2011 using a Speedstar SS15 rotary rig. Continuous core samples were collected from 324 to 411 m BG at K2B using a 3.04 m × 75 mm core barrel with polyvinyl chloride (PVC) liners.
 At K2A, subsamples (∼75 mm long) were collected from continuous cores at 1 m intervals (n = 290) in the field. Immediately after retrieval, the outer 2–5 mm of each core sample was removed to minimize contamination from the drill mud. All samples were placed in medium-sized Ziploc™ polyethylene resealable bags, and the atmospheric air was squeezed out of the bags prior to sealing. Each bagged sample was then placed inside a larger-sized Ziploc polyethylene resealable bag, with the air again squeezed out prior to sealing. The bagged cores were placed in coolers and kept at ambient surface temperature (∼5–10°C) until they could be transported to the University of Saskatchewan where they were kept in the coolers at room temperature until analysis.
 In contrast to the sample collection used for the core samples at K2A, the entire length of core from each core barrel at K2B remained in the PVC liners. Once the drill water drained from the tube, the ends of the PVC tubes were capped and sealed with duct tape, and the core tubes were shipped to the laboratory where they were stored for 2 days before subsampling. In the laboratory, the PVC tubes were cut lengthwise using a saw. After a core was removed from the tube, the sampling procedure was as described earlier. Core samples were collected at 1 m intervals from 324 to 411 m BG (n = 85).
 Bulk density (ρ) and particle density (Sg) were determined on core samples collected at 10 m depth intervals at K2A (n = 33). Bulk densities were determined using ASTM D7263-09 [American Society for Testing and Materials (ASTM), 2009], and Sg was determined using ASTM D854-10 [ASTM, 2010]. These data were used to calculate total porosity (nT) of the samples.
 Core samples (n = 375) were analyzed for the stable isotopes of water using H2O(liquid)-H2O(vapor) equilibration on a Picarro L1102-i isotopic water liquid analyzer. In the laboratory, the medium-sized Ziploc bags containing the samples were inflated with H2O-free dry air and resealed. Each inflated bag was returned to the large Ziploc bag from which it came and the air from the large bag removed before resealing. The samples were allowed to equilibrate isothermally to 100% relative humidity at room temperature for 7 days prior to analysis. Testing (not presented) showed that this equilibration time provided optimum δ18O and δ2H results. Two water standards with δ18O and δ2H values that bracketed that of the pore waters in the core samples were prepared and run after every four samples to correct for instrument drift and to normalize the results to the Vienna Standard Mean Ocean Water/Standard Light Antarctic Precipitation (VSMOW/SLAP) scale. Details of the core sampling, sample preparation, and the analytical methods used are presented by Hendry et al. [2011a, 2011b] and Wassenaar et al. . The accuracy of the analytical method, based on laboratory standard waters and the analysis of replicate core samples, was better than ±0.3‰ for δ18O and ±0.8‰ for δ2H relative to VSMOW.
 Selected core samples (K2A: 41, 80, 98, 120, 138, 160, 187, 204, 220, 260, and 320 m BG; K2B: 337, 350, 353, 384, 389, 405, and 409 m BG) were squeezed in a high-pressure mechanical squeezer to obtain pore water samples for chemical and isotopic analyses. The samples used for squeezing were chipped and immediately packed as tightly as possible into the squeeze cylinder (316 L stainless steel; 50 mm diameter × 80 mm long). The piston was inserted in the cylinder and placed in the hydraulic press where the pressure was increased to 50 MPa. This pressure was selected for squeezing because testing showed no measurable effects on the δ2H and δ18O values (A. L. Bangsund et al., Effects of incremental squeezing on the chemistry and stable isotopes of pore waters of high-porosity and consolidated argillaceous drillcores, submitted to Clay Minerals, 2013). The samples were maintained under pressure for 3–5 days. During squeezing, pore water passed through a 0.45 µm stainless steel filter before exiting via a port at the base of the cylinder where it was collected in a clean 60 cm3 syringe. The pore water samples, which ranged in volume from 3 to 10 cm3, were subsequently transferred to 20 mL scintillation vials and tightly capped until isotopic analysis. Pore water samples collected from the squeezed samples were analyzed for δ2H and δ18O using laser absorption spectroscopy using the method described by Lis et al. . The accuracy and precision of this method, based on laboratory standard waters and the analysis of replicate core samples, were better than ±0.1‰ for δ18O and ±0.8‰ for δ2H relative to VSMOW reference.
3.3. Pore Water Sampling and Analyses
δ18O and δ2H values from monitoring wells installed in the Mannville at K2A and K2B and at eight other sites installed by Mosaic within 10 km of K2A (maximum and minimum screen depths of 392 and 483 m BG, respectively) were sampled between April and October 2011. Analytical methods for these δ18O and δ2H measurements were as described earlier. Water samples from seepage faces were also collected from the K1 (n = 31) and K2 (n = 55) mine shafts by Mosaic between 1985 and 2004.
3.4. Assessment of Core Contamination by Drilling
 Contamination of the core subsamples (both vapor and squeezed samples) by drilling fluid was determined by spiking the drill mud (water was used as a drill fluid at both sites) with 99% D2O prior to drilling to yield δ2H values of −73‰ to +112‰ and −3‰ to +90‰ at K2A and K2B, respectively. Drill fluid samples were collected routinely during coring for analyses of δ2H and δ18O. These values were compared to measured values from the core and squeezed samples to evaluate fluid contamination. Details and results are provided in the supporting information (and associated Figure S1). Samples that showed isotopic contamination from drill mud by isotopic offsets from the meteoric H and O isotope corelationship were removed from the analytical database.
4. Results and Discussion
4.1. Total Porosity
 The nT values for the shale are uniform with depth and have a mean of 0.33 ± 0.04 (n = 37). These values are consistent with measured values from the Lower Colorado (mean = 0.33 ± 0.05; n = 36; E. E. Schmeling and M. J. Hendry, unpublished data, 2013) 160 km north of Saskatoon, Saskatchewan (Figure 1) and a value of 0.34 for the Pierre Shale in South Dakota reported by Neuzil . Limited data for the Mannville and till yield mean nT values of 0.34 ± 0.3 (n = 2) and 0.24 (n = 1), respectively, consistent with values (mean = 0.27 ± 0.03; n = 16) from the Mannville measured by Schmeling and Hendry (unpublished data, 2013).
4.2. Stable Isotopes of Water
 A cross plot of the δ18O and δ2H values for pore water from shale core samples (vapor equilibration and squeezing) is presented in Figure 4a. These data sets yield a linear trend (δ2H = 5.31δ18O − 44.7, R2 = 0.60), identified as the K2 Shale Trend. The shallow slope of the Shale Trend with respect to the local meteoric water line (LMWL) for Saskatoon (δ2H = 7.73 δ18O − 1.72, R2 = 0.96; average weighted mean values of −111‰ and −14.1‰ for δ2H and δ18O, respectively) [Hendry et al., 2011a] suggests these samples have been affected by evaporation or mixing with older, evaporated water. The δ18O and δ2H values from mine-shaft seepage waters collected from the Mannville to the Duperow at the K2 mine also plot on or close to the Shale Trend (Figure 4b), and thus these values and their origin and evolution may be related to the Cretaceous shale data.
 Unlike the evaporitic trend in the shale and data from deeper in the basin, the δ18O and δ2H values of groundwater samples from the till and Mannville are more depleted and plot on the LMWL (Figure 4b), suggesting an unaltered meteoric origin for both waters. In the case of the till waters, the depleted meteoric values are in keeping with observations of a colder glacial age origin made at other sites across the Interior Plains of North America [cf. Hendry et al., 2011a; Remenda et al., 1994]. The presence of depleted meteoric water values from seepage inflows and wells completed in the Mannville is consistent with Grasby and Betcher  and Grasby and Chen , who suggest that depleted meteoric water intruded far into the Mannville as subglacial meltwater during Pleistocene glaciation. Alternatively, more local injection of meltwater may have occurred through collapse features [Wittrup and Kyser, 1990] or faults through the Cretaceous associated with dissolution of the underlying Prairie Evaporite [Gendzwill and Stauffer, 2006].
4.3. The δ18O-Depth Profile
 The δ18O-depth profile for the Cretaceous is presented in Figure 2 and used for interpreting depth trends in the isotopic data. This degree of isotopic resolution would not have been evident with less frequent sampling intervals (e.g., using piezometers). The high-resolution isotopic profile allowed us to better interpret distinctive changes in the profile through formations and at the upper and lower boundaries of the profile (important for transport modeling), as well as demonstrate goodness of agreement with data from core holes K2A and K2B.
 The δ18O values decrease from −13‰ at 3 m BG in the till to between −15‰ and −17‰ between 27 and 34 m BG in the Pierre (16–23 m below the till-shale contact). Below 35 m BG, the δ18O values increase to between −12‰ and −9‰ in the Second White Speckled Shale (264 and 274 m BG). At depths greater than 274 m BG, the δ18O values decrease to between −16‰ and −19‰ at the top of the Mannville (408 and 405 m BG). The good agreement between δ18O values from cores collected near the bottom of K2A (321 m BG) and samples collected from the top of K2B (324 m BG) provides support for the assumption that results from the core profiles can be combined. It further suggests that the isotopic vertical variability in the shale profile is likely to be minor over distances of at least 10 km (distance between core holes).
 The δ18O values for vertically adjacent shale samples are typically ±1‰ (Figure 2). This range in values exceeds the precision of the analytical method (±0.3‰) for δ18O and suggests a potential analytical limitation with respect to obtaining more precise isotopic values from shale samples; however, this observation does not alter the trend in the δ18O values with depth or its interpretation. The poor precision may be attributed to interference by natural organic compounds that absorb in the same wavelength range as the isotopologues of water [Hendry et al., 2011b].
 Superimposed on the well-defined depth trend in δ18O values, the high-resolution profile shows a discontinuity in δ values of about 1.5‰ in the Second White Speckled Shales (256–281 m BG). The values decrease and then increase through this zone. The cause of this shift is not clear, but we believe it is caused by interferences related to elevated concentrations of CH4 gas in this unit in keeping with testing by Hendry et al. [2011b]. Investigations into effects of dissolved gases on the stable isotopes of water are ongoing.
 The δ18O values of squeezed samples are generally consistent with the core equilibration values. However, δ18O values from squeezed samples from the Pierre shale are slightly greater than those from core equilibration (about +1‰), with one sample (220 m BG) about +2‰ greater than the corresponding core equilibration samples. The cause of the differences is not clear but may again be due to organic interferences or exchange and release of structured water associated with the clay minerals or bound to mineral surfaces.
 The δ18O values from the Mannville aquifer (seepage inflow samples and Mannville well samples) yield a uniform value of about −16.1‰ across the study area. These values are consistent with the core equilibration and squeezed data from the overlying shale and further support the depth trend in the δ18O values in the shale.
 The shape of the δ18O profile through the shale and into the Mannville aquifer suggests that the 18O diffused from the shale downward into the aquifer and upward toward the top of the shale. The extent of the upward and downward trends in data from the Second White Speckled Shale (about 256 m downward and about 395 m upward) suggests that this profile developed over a considerable period of time and in the case of the increasing negative values with depth across the lower 125 m, implies that this trend was the result of the invasion of a different fluid into the aquifer. As stated earlier, the depleted meteoric values in Mannville are consistent with either Grasby and Betcher  and Grasby and Chen  or Gendzwill and Stauffer  and Wittrup and Kyser  who respectively suggest that meteoric water intruded down-dip far into the Mannville as subglacial meltwater during Pleistocene glaciation or alternatively intruded through local collapse features or faults during glaciation. The shallow extent of the deviation in the δ18O profile in the upper shale from a negative upward trend to a positive upward trend extending into the till suggests downward migration of 18O through this zone occurred recently. This depth trend and the δ18O values in the till are consistent with profiles that develop with depth below water tables in surficial aquitards throughout the Interior Plains Region of Canada after the onset of the Holocene [cf. Hendry and Wassenaar, 1999; Hendry et al., 2011a].
4.4. Defining Initial Conditions for Transport Modeling of δ18O in the Cretaceous
 The flushing of the aquifer with isotopically depleted meteoric water during Pleistocene glaciation is termed aquifer “activation” in the current paper. To simulate the development of the δ18O profile in the shales and Mannville aquifer since activation, knowledge of the δ18O profile in the Cretaceous and the underlying formations prior to activation of the Mannville was required.
4.4.1. Defining Present-Day Conditions in the Sedimentary Sequence
 The present-day isotopic profile was generated by combining the δ18O values for samples of inflow water in the K2 shafts with the depth profile through the Cretaceous. The composite isotopic profile is presented in Figure 3. The δ18O-depth trend from the top of the Cretaceous to the base of the Duperow Formation shows increasing values with depth punctuated by the negative excursion centered on the Mannville aquifer (described earlier). The meteoric values in the Mannville (based on measurements of well waters, core analyses, and seepage inflow measurements; about −16.5‰) continue through the underlying Lodgepole and into the Nisku. The available data do not clearly point to meteoric values in the Lodgepole and Nisku resulting from downward seepage along the mine-shaft walls from the overlying Mannville or indicate if the measurements reflect conditions in these units. Below this zone, the δ18O values increase to −0.5‰ to −2‰ in the Duperow.
 Excluding the negative excursion, the data show a near-uniform increase with depth in the δ18O values from the upper 260 m of shale through the carbonates to the base of the Duperow (Figure 3). The increasing stable isotopic composition of formation fluids with increasing depth in the WB is consistent with observations by Hitchon and Friedman , Kyser and Kerrich , and Wittrup and Keyser , who attribute this trend to mixing between two waters: meteoric water at the top of the sequence and evolved basin brines at the base. The lower end member of evolved brine is probably located in or below the Prairie Evaporite. Extrapolating the near-linear trend for the δ18O values from the Cretaceous shale, through the Duperow to the Prairie Evaporite, yields an end member, and lower boundary, value of about +5‰. This value is similar to maximum δ18O values reported for the WB (+4.6‰ to +6.9‰; Figure 4a) [Rostron and Holmden, 2000].
4.4.2. Estimating the δ18O Profile Prior to Activation of the Mannville Aquifer
 The δ18O-depth profiles (preactivation and postactivation of the Mannville) were generated by modeling the vertical migration of 18O using the 1-D advective-diffusive transport equation:
where De is the effective diffusion coefficient, V is the average linear velocity, C is the mass concentration of the solute, z is distance, and t is time. In equation (1), De is defined according Fick's first law as
where Jd is the diffusive mass flux rate and ne is the effective porosity. The finite element model CTRANW [Geo-Slope International Ltd., 2008] was used to simulate transport using equation (1).
 The values of De and ne for each geologic unit were estimated based on the results of previous laboratory and field measurements. The parameter ne is reasonably approximated by nT for stable isotopes of water [Hendry and Wassenaar, 1999; van der Kamp et al., 1996]. Consequently, ne values for δ18O in the shale, till, and Mannville were assigned the mean measured values of 0.33, 0.24, and 0.34, respectively. The nT (and ne) for the carbonates was estimated at 0.1 to reflect a likely reduced ne value in massive carbonate deposits.
 The De values for the δ18O in the till, shale, and Mannville aquifer were assumed to be 2.3 × 10−10 m2/s and thus consistent with De measurements for δ2H in till and Upper Cretaceous shale by Hendry and Wassenaar  and in the case of the till, the same as that calculated by Hendry et al. [2011a]. Further, because these units have similar nT values with no indication of pore structure, use of a constant De values was not unreasonable. The De values for the δ18O in the carbonates were calculated from empirical relationships between De and ne as reported by Boudreau , Boudreau and Meysman , and Hendry et al. .
 The upper and lower oxygen isotope end members in Figure 3 (i.e., −16.5‰ and +5‰) were used as the upper and lower boundaries in this modeling exercise. Because the shales were deposited in a marine environment, initial conditions throughout the sedimentary sequence were assumed to be that of present-day seawater (0.0‰). This assumption is consistent with initial conditions used in other marine-deposited aquitards [Mazurek et al., 2011] but requires testing in the case of the underlying carbonates. Note also that only diffusive conditions were simulated (i.e., V = 0 for all cases). See section 4.5 for simulations in which advection was considered.
 The simulated δ18O profiles provide a profile of initial conditions (prior to activation) that appears reasonable when compared to the measured profile in the shale and underlying carbonates (Figure 3). The inflection point in the simulated data at the base of the Mannville reflects the differences in ne values used for the mesoclastic versus the Paleozoic carbonates. The time to generate the measured profile using these model parameters is difficult to estimate from the simulations but brackets 15–25 Ma, with 20 Ma yielding the best overall fit to the data (Figure 3; Note: other evolution times were simulated but not presented). For comparison, simulations performed by assigning the ne and De values for the shale to the carbonates also yield reasonable fits to the profiles after 20 Ma evolution time (data not presented). Similarly, the use of an initial condition other than 0.0‰ for the carbonates or using an end member of −3.0‰ at the top of the Duperow yields an equally good fit, but with different evolution times. The simulations provide assurance that the profile used for the initial conditions in the shale and carbonates prior to activation of the Mannville is reasonable; however, the evolution times are unlikely appropriate given the simplistic nature of the model. For example, the influences of longer-term geologic events, such as the extensive erosion of up to 3 km of Tertiary and Upper Cretaceous sediments during the Tertiary over the past 56–2 Ma, is not incorporated into the model. Incorporating these mechanisms may provide an improved fit between the simulated profile and the measured data; however, the profile as shown in Figure 3 is believed to provide a reasonable initial condition for the δ18O profile prior to activation of the Mannville aquifer.
4.5. Characterizing the Evolution of the δ18O After Activation of the Mannville Aquifer
 Defining the activation history of the Mannville aquifer is based on a number of assumptions with respect to the timing and magnitude of isotopic values in the aquifer before and after activation(s). The most commonly used approach to determine the time to generate tracer profiles resulting from the activation of an aquifer is to fix the concentration in the aquifer at the present-day measured values and allow the profile to propagate into or out of the aquitard(s) for given time periods [cf. Mazurek et al., 2011].
 The δ18O profile in the shale and the underlying carbonates was simulated using this approach for a range of elapsed times since activation using the present-day δ18O value of −16.5‰ in the Mannville. For these simulations, activation times for the Mannville were selected to coincide with the onset of major glacial periods (Figure 5). Once the aquifer was activated, it remained so to present day (i.e., perpetual flushing). In keeping with the preactivation modeling, the De and ne values in these simulations were as described earlier. The exception was the De value for the Mannville, which was increased to 2.3 × 10−9 m2/s to approximate the effect of dispersive mixing arising from lateral groundwater flow within the Mannville. Also in keeping with the earlier modeling, the mode of solute transport in the model was maintained as pure diffusion (V = 0).
 Good fits between the simulated and measured δ18O values (especially in the lower 150 m of the shale) were obtained for elapsed times of between 0.2 and 0.5 Ma since activation of the Mannville aquifer (Figures 5a and 5b). Additional simulations (not presented) using preactivation conditions generated after 25 Ma (Figure 3) yield very similar profiles to those presented in Figure 5 and the same range in activation times for the Mannville aquifer.
 Independent of the activation of the Mannville aquifer, the δ18O value at the water table in the till was changed from −16.5‰ to the measured value of −13.0‰ (see Figure 2) at 10 ka BP. This time was selected to reflect the well-constrained onset of the Holocene in Saskatchewan [Hendry and Wassenaar, 1999, 2011; Hendry et al., 2011a]. In these simulations, stipulating the onset of the Holocene at 10 ka yields a good fit to the measured data in the upper 25 m of the profile (all simulated profiles yield identical results, as expected; Figure 5b), supporting the findings from other near-surface sites in Saskatchewan (see references earlier).
 Although these simulations provide good overall fits to the measured δ18O profile, no realistic physical mechanism, other than perpetual flushing of the Mannville, would be represented by a constant value throughout the aquifer. Therefore, a more realistic assumption for the activation of the Mannville could be represented by one or more flushings associated with individual glaciations. The potential of these glaciations were modeled by assuming that the meteoric water that flushed through the aquifer by glacial loadings caused a near-instantaneous change in the δ18O value within the Mannville. After activation, the δ18O values in the aquifer gradually increase as diffusion into the aquifer from overlying and underlying formations occurred. In these simulations, the upper and lower boundary conditions and the De and ne values of the formations were the same as used in Figure 5a, and diffusion was again assumed to be the only transport mechanism (V = 0). An instantaneous activation value of −20.0‰ was used in the Mannville to be consistent with Pleistocene waters discharging in springs from the basin in west-central Manitoba [Grasby and Chen, 2005]. The instantaneous activation was applied to the Mannville for a range in single and multiple glaciations ranging from the fourth Kansan (0.5 Ma BP) to the Wisconsin (20 ka BP; Table 1) to assess the impact of Pleistocene glaciations on the measured profile.
Table 1. Glacial Cycles Applied to the Mannville Aquifer in the Transport Modeling Scenariosa
Glaciation Cycles Included
Run 1 (not presented) was the preactivation transport simulation.
 A comparison of fits between the measured and simulated δ18O profiles for flushing of the aquifer during these glacial events shows the best fit was obtained for activation as a single flush during the third Kansan glaciation (0.3 Ma BP; Figure 6). However, good fits are also obtained for a number of other glacial events that included multiple flushings of meteoric water through the aquifer since the first Kansan (0.5 Ma) and for scenarios in which flushing of the aquifer occurred during the second Kansan to the fourth Kansan (approximately 0.4–0.2 Ma BP; Figure 6). These simulations show that the most recent glaciations (Illinois and Wisconsin, 135 and 20 ka BP) would not reproduce the measured profile (due to the similarity in the profiles, the multiple flushes are not presented in Figure 6).
 There is also a reasonable possibility that flushing of the Mannville (to glaciogenic values of −20.0‰) was not completely efficient. Consequently, the δ18O values in the aquifer following activation would be a mixture of meteoric and residual aquifer waters. Considering the K2 Shale Trend, however, the presence of residual water would have pushed the Mannville samples away from the Meteoric Water Line, which is not observed (Figure 3b). Nonetheless, simulations of partial flushing were performed by setting the δ18O value of the Mannville following activation to −18‰. The same upper and lower boundary conditions, initial conditions (20 Ma), and diffusive transport conditions were as described previously. The activation sequencing presented in Table 1 was then repeated. The results of the partial concentration scenarios (data not presented) extend the best fit simulations to include the Illinois glaciation (0.1 Ma) when coupled with Kansan glaciations. By increasing the reset values and coupling mid-Kansan to Illinois glaciations (0.3–0.1 Ma), a better fit to the data is also obtained. These simulations show that the profiles in the shale and Mannville aquifer can be generated via a partial resetting of the values in the Mannville during glaciations from the first Kansan to the Illinois. Additional knowledge of the reset value(s) is required to define which glaciation or combination of glaciations results in the measured profile. As such, no further 1-D simulations were warranted. The 2-D transport modeling across the basin may provide additional knowledge of the reset values and timing of glacial events.
 Consistent with the perpetual activation simulations (Figure 5), the δ18O value at the water table in the till was changed from −16.5‰ to the measured value of −13.0‰ (see Figure 2) at 10 ka BP. These simulations yield good fits to the measured data in the upper 25 m of the profile (Figure 6b), again supporting the findings from other near-surface sites in Saskatchewan.
4.6. Effects of Velocity on the Evolution of the δ18O After Activation of the Mannville Aquifer
 A final set of simulations was undertaken to explore whether differences in the best fit simulations might occur if a vertical component of flow across the shale (advective transport) was included. These simulations were conducted in consideration of the potential effects due to large hydraulic gradients that might have been imposed during glacial periods [Bense and Person, 2008; Nasir et al., 2011]. The initial conditions were based on the values through the geologic profile after 20 Ma of development and an instantaneous activation δ18O value of −20.0‰ for the Mannville (consistent with Figure 6). Two sets of simulations were performed with upward and downward components of groundwater flow, respectively, applied across the shale. The inclusion of a downward component of advection was consistent with present-day measured head data in the aquitard [Smith et al., 2013] and what might be expected during glaciation. Although not representative of present-day conditions, the addition of upward advection would account for periods when the Mannville aquifer was activated by lateral fluid migration or local injection down collapse or fault features during glacial advance. Based on Darcy calculations conducted using the measured K and n values for the shale and considering that the bulk K might be as much as 0.5 orders of magnitude greater than the sample measurements, estimates of groundwater velocity could potentially exceed 250 m/ka under 2000 m of hydraulic head. In these simulations (i.e., for both upward and downward advection), groundwater velocities were assumed to range broadly from 0.01 to 10.0 m/10 ka. The upward limit of the range was set to 10.0 m/ka to represent an average of periods of very high gradient interspersed with periods of low to negligible gradient.
 The simulated profiles for downward V values of 0.01, 0.05, 0.1, and 10.0 m/10 ka and measured δ18O values are presented in Figures 7a–7d. The best fit simulated profiles are those for V values of 0.05 m/10 ka and less. The simulations for greater V values (0.1–10.0 m/10 ka) yield progressively poorer fits to the measured data, to the point at 10 m/ka where no fit is possible. These results suggest that the V of the aquitard has been <0.05 m/10 ka since activation of the aquifer, which in turn implies that either (1) the bulk K of the shale is substantially less than the K measured using the core samples, (2) the periods of high hydraulic gradients during glaciation were short and infrequent, (3) mechanical loading significantly reduced n during glaciation, (4) the base of the glaciers was frozen (did not impart a hydraulic connection), or (5) some combination of these considerations. Note that if we use the lower of the measured values determined from the core samples (2 × 10−12 m/s), the estimate of average n (0.33), and back calculate an average hydraulic gradient from the maximum possible velocity (0.05 m/ka), a surplus hydraulic head of only 10 m is determined. This illustrates the importance of the estimate of K, as this hydraulic head value seems remarkably low considering the extended periods of glaciation experienced by the area.
 The fit between the simulated and measured data for the till and upper shale for δ18O shows that V (to the extent tested) has a negligible effect on the estimated timing for the onset of the Holocene, with all cases providing a reasonable fit for the timing of the onset at about 10 ka (data not presented). These findings are consistent with those of Hendry and Wassenaar .
 Results for the simulated profiles for upward V values of 0.01, 0.05, 0.1, and 10.0 m/10 ka) plotted against measured data are presented in Figures 7e–7h. As was the case for the simulations of downward V, the best fits are obtained for very low upward average velocities. The higher velocity case significantly departs from the measured values in the shale. Similar to the case for downward advection, numerous reasons may explain the minimization of upward advective migration, most of which are suggested earlier. Based on these results, we can surmise that if multiple periods of Mannville activation occurred, they were not accompanied by significant long-term vertical hydraulic gradients.
5. Summary and Conclusions
 High-resolution profiles of the isotopes of pore water (δ18O and δ2H) were collected from 400 m of Cretaceous shale in the Williston Basin. These high-resolution profiles exhibit well-defined trends with depth, from ground surface through the surficial till (12 m thick) and Cretaceous shale and into the underlying Cretaceous Mannville aquifer (100 m thick) and show that the hydrogeology of the basin cannot be considered static over geologic time. Pleistocene glaciations likely modified the groundwater flow regime in the basin by injecting large volumes of isotopically depleted δ18O and δ2H meteoric waters into the Mannville either from the glacial recharge area over more than 180 km to the east or from nearby collapse or fault features that penetrate the shale. Because these glacially introduced waters have unique isotopes signatures (cold-climate meteoric recharge), they can function as long-term, basin-scale tracers. The distribution (shape and vertical extent) of the tracers in the shale is consistent over at least 10 km (distance between shale core holes and wells in the Mannville). These profiles allow us to begin to develop testable hypotheses for the paleohydrogeology of the basin.
 The shale data were used in conjunction with detailed isotopic data from the underlying carbonate sediments (400 m thick) to define the present-day vertical δ18O profile through about 800 m of sediment in the basin (including the shale). This present-day profile was used to estimate initial conditions through the sedimentary sequence prior to the activation of the Mannville aquifer by glaciation(s) via modeling. Modeling of the activation of the aquifer and the resulting shale profile shows that the profile can be simulated by diffusive transport only. Modeling suggests that the vertical groundwater velocity (either upward or downward) averaged <0.05 m/10 ka over the past 0.24–0.43 Ma. This suggests that the significant hydraulic gradients imposed during periods of glaciation resulted in very little advection. Considering the estimated K of the shale, the exact mechanisms that led to the minimization of advection are not obvious and must be explored in greater depth. The modeling further suggests that the glacial recharge events occurred during one or more glacial periods, although the exact timing(s) cannot be determined from the available data. The stable isotope profiles through the surficial till and upper shale are also consistent with the timing of the onset of the Holocene (10 ka) across the Interior Plains of North America as reported by other researchers.
 The first application of extremely high-resolution stable isotopic profiles through the shale aquitards facilitated observation and testing of model hypotheses with respect to solute transport over scale and timeframes that would not have been possible using conventional widely spaced hydrogeologic installations. The measured changes in the isotopic profiles allowed us to comment on the level of accuracy of the measured values, changes in the profiles through individual formations and at the upper and lower boundaries of the profile (important for transport modeling), and goodness of agreement between data from core hole K2A and K2B. In the future, more accurate estimates of the timing of the glacial recharge event(s) in the Mannville might be possible by combining the findings of the stable isotope profiles (presented here) with similar profiles for one or more additional conservative tracers and/or comparison with additional vertical profiles closer to or further from the recharge area or source area of the glaciogenic fluid and by interpreting these multiple tracer profiles using a 2-D transport model of the basin.
 V. Chostner, L. Smith, and F. Nelson provided assistance with data collection and processing; E. Schmeling and K. Relland assisted with transport modeling; and G. Koehler shared his insights into the stable isotopes of the WB. Funding was provided by NSERC through their IRC and CRD programs and by the Mosaic Company (M.J.H.). The Mosaic Company also provided access to wells for water sampling and the analyses from mine-shaft seepage samples.