Groundwater samples were collected from water table wells and piezometers (hereafter both are referred to as wells) installed at both sites (Table 1). At CFO1, groundwater samples were collected from six individual water table wells (DMW1, DMW2, DMW3, DMW4, DMW5, DMW6) and eight sets of nested wells with one well screened at the water table and one well screened 20 m below ground (BG) (DP10-2 and DP10-1, DMW10 and DP11-10b, DMW11 and DP11-11b, DMW12 and DP11-12b, DMW13 and DP11-13b, DMW14 and DP11-14b, DMW15 and DP11-15b, and DMW16 and DP11-16b). Wells DP10-2 and DP10-1 were located directly adjacent to the EMS on the hydraulically down-gradient side. At CFO4, groundwater samples were collected from eight water table wells (BC1, BC2, BC3, BC4, BC5, BMW1, BMW3, BMW7) and four sets of nested wells, with wells screened across the water table and at 15 m BG. Two of these nests were located adjacent to the EMS (BMW2 and BP10-15e, BMW4 and BP10-15w) and two were hydraulically down-gradient of the EMS (BMW5 and BP5-15, BMW6 and BP6-15).

Groundwater samples were collected for ion analysis (Cl- and N species) quarterly between April 2010 and August 2015. All water samples were collected using a bailer after purging (1–3 casing volumes) and stored at ≤4 ∘C prior to analysis. Samples for δ15NNO3 and δ18ONO3 were collected from wells at CFO1 on 1 January and 1 May 2013. Samples for δ15NNO3 and δ18ONO3 at CFO4 were collected on 27 October 2014. Wells were purged prior to sample collection (1–3 casing volumes), and samples filtered into high-density polyethylene (HDPE) bottles in the field and frozen until analysis.

Hydraulic heads in monitoring wells were determined using manual measurements (approximately monthly, 2010–2015). Hydraulic head response tests were conducted on the majority of the wells at the sites to determine hydraulic conductivity (K) of the formation media surrounding the intake zone. These tests were either a slug test (water level decline after water addition) or bail test (water level recovery after water removal) depending on the location of the water level within the well at the time of testing. K was determined from the hydraulic head responses using the method of Hvorslev (1951).

A continuous core was collected at CFO1 immediately adjacent to well DP11-13b on 1 May 2013 (Fig. 1). Additional core samples were collected from 1 to 5 June 2015 along a transect hydraulically down-gradient of the southeastern side of the EMS at CFO1, where hydrochemistry data suggested leakage from the EMS (see Sect. 3). During this 2015 drilling campaign, core samples were collected at four locations (DC15-20, DC15-21, DC15-22, DC15–23) to depths of up to 15 m below surface and distances of up to 100 m from the EMS between wells DMW3 and DP11–14.

Continuous core samples were retrieved using a hollow stem auger (1.5 m core lengths) with 0.3 m sub-samples collected at approximately 1 m intervals ensuring that visually consistent lithology could be sampled. Core samples for Cl- were stored in ZiplocTM bags and kept cool until analysis. Core samples for N-species analysis were stored in Ziploc bags filled with an atmosphere of argon (99.9 % Ar) to minimise oxidation and kept cool until analysis. Subsamples of each core (250–300 g) were placed under 50 MPa pressure in a Carver Auto Series NE mechanical press with a 0.5 µm filter placed at the base of the squeezing chamber, which was placed within an Ar atmosphere to minimise oxidation. A syringe was attached to the base of the apparatus and 15 mL of filtered pore water were collected for analyses within 3.5 to 6.0 h (Hendry et al., 2013).

Samples of liquid manure slurry were collected directly from the EMS at both sites and the catch basin (containing local runoff from the feedlot) at CFO1 using a pipe and plunger apparatus to sample from approximately 0.5 m below the surface. The slurry collected was subsequently filtered (0.45 µm) to separate the liquid and solid components. The water filtered from samples collected from the EMS or catch basin is hereafter referred to as manure filtrate.

Nitrate in groundwater that has undergone denitrification is commonly reported as being identified by enrichment of δ15NNO3 and δ18ONO3 with a slope of about 0.5 on a cross-plot (Clark and Fritz, 1997). However, published studies of denitrification in groundwater report slopes of up to 0.77 (Mengis et al., 1999; Fukada et al., 2003; Singleton et al., 2007). The relationship between isotopic enrichment of 15NNO3 and 18ONO3 and the fraction of NO3-N remaining during denitrification can be described by a Rayleigh equation: R=R0fd1β-1, where R0 is the initial isotope ratio (relative to the standard) of the NO3- (δ18ONO3 or δ15NNO3), R is the isotopic ratio when fraction fd of NO3- remains, and β is the kinetic fractionation factor (>1) (Böttcher et al., 1990; Clark and Fritz, 1997; Otero et al., 2009; Xue et al., 2009). Kinetic fraction effects are commonly also expressed as the enrichment factor, ε=11000(β-1) . In the case of a constant enrichment factor, fd can be calculated from measured δ15NNO3 (or δ18ONO3), if the initial δ15NNO3 (δ15N0) is known; fd=exp⁡δ15NNO3-δ15N0ε. The fraction of NO3-N removed from groundwater through denitrification is then given by (1-fd). The concentration of NO3-N that would have been measured if mixing was the only attenuation mechanism (NO3-Nmix) can also be calculated by dividing the measured concentration by fd.

A subset of 20 samples with isotopic values of NO3- indicative of denitrification were identified, and for each of these samples fd (mean and standard deviation) was calculated from Eq. (2) using a Monte Carlo approach with 500 realizations. The distribution of ε values was defined based on measured data. If the initial δ15NNO3 is known, ε for δ15NNO3 (ε15N) can be determined from the slope of the linear regression line on a plot of ln⁡(fd) vs. δ15NNO3 (Böttcher et al., 1990). If the initial δ15NNO3 and fd are not known, as is the case here, ε15N can be determined from the slope of the regression line on a plot of ln⁡(NO3-N) vs. δ15NNO3, which will be the same as on a plot of ln⁡(fd) vs. δ15NNO3. In situ variations in temperature and reaction rates may affect the enrichment factor (Kendall and Aravena, 2000) and this was accounted for by allowing for variation in ε15N within the Monte Carlo analysis. The enrichment factor for δ18ONO3 (ε18O) was calculated by multiplying the δ15NNO3 by a linear coefficient of proportionality determined for each CFO from the slope of the denitrification trend on an isotope cross-plot (see Sect. 3.2).

For each realisation, initial isotopic values (δ15N0 and δ18O0) were determined by Excel Solver such that the difference between fd calculated from δ15NNO3 and δ18ONO3 was minimised (<1 % difference). The ranges of δ15N0 and δ18O0 were limited based on measured data and literature values (see Sect. 3.2). This approach neglects the effect of mixing of groundwater with differing isotopic values and is valid if the concentration of NO3- in the source is much greater than background concentrations such that the isotopic composition of NO3- is dominated by the agriculturally derived end-member.

A binary mixing model that also accounts for decreasing NO3-N concentrations in response to denitrification was used to quantify NO3- attenuation by mixing and estimate the initial concentrations of Cl- and NO3-N. The measured concentration of Cl- was assumed to be a function of two end-members mixing, described by Cl=fmCli+1-fmClb, where Cl is the measured concentration of Cl- in the groundwater sample, Cli is the concentration of Cl- at the initial point of entry of the agriculturally derived NO3- to the groundwater system, Clb is the concentration of Cl- in the background ambient groundwater, and fm is the fraction of water in the sample from the source of agriculturally derived Cl- (and NO3-) remaining in the mixture.

The concentration of NO3-N was also assumed to be a function of two end-members mixing but with an additional coefficient, fd (the fraction of NO3-N remaining after denitrification), applied to account for denitrification. The measured NO3-N concentration was thus described by NO3-N=fdfmNO3-Ni+1-fmNO3-Nb, where NO3-N is the concentration of NO3-N measured in the groundwater sample, NO3-Ni is the concentration of NO3-N in the source of agriculturally derived NO3- at the initial point of entry to the groundwater system, and NO3-Nb is the concentration of NO3-N in the background ambient groundwater. This mixing calculation was only conducted on samples for which NO3- dominated total-N (NH3-N < 10 % of NO3-N) so that nitrification of NH3 could be neglected.

If Cli is much greater than Clb and NO3-Ni is much greater than NO3-Nb, then fm is insensitive to background concentrations and these terms can be neglected (see Sect. 4.2 for further discussion of this assumption). In this case, Eqs. (3) and (4) reduce to Cl=fmCli,NO3-N=fdfmNO3-Ni. Solving Eq. (6) for fm and substituting into Eq. (5) yields NO3-NiCli=1fdNO3-NCl. Thus, for each groundwater sample, the ratio of NO3-N/Cl- at the initial point of entry of the agriculturally derived NO3- to the groundwater system NO3-NiCli can be simply calculated using measured concentrations, and fd estimated from NO3- isotope data. This provides a relatively simple method to identify agriculturally derived NO3- from different sources (e.g. EMS vs. manure piles) if they have different NO3-N/Cl- ratios. Estimated Cli and NO3-Ni are reported as the mid-range value with uncertainty described by the minimum and maximum values. These initial concentrations are at the water table for top-down inputs, or at the saturated point of contact between the EMS and the aquifer for leakage from the EMS. This analysis assumes that a sampled water parcel consists of water with agriculturally derived NO3- that entered the aquifer from one source at one point in time and space and has since mixed with natural ambient groundwater. Any NO3- produced during nitrification after the anthropogenic source water enters the aquifer is implicitly included in NO3-Ni. The error in NO3-NiCli- was assumed to be dominated by error in the estimated fd, with the measurement error in NO3-N and Cl- considered negligible.

The initial concentrations of the agriculturally derived NO3- source (NO3-Ni and Cli) were estimated by simultaneously solving Eqs. (5) and (6) using Excel Solver (GRG nonlinear). The absolute minimum values of NO3-Ni and Cli were defined by measured concentrations (e.g. if Cli=Cl, fm=1). Maximum values of NO3-Ni and Cli were defined based on measured concentrations of NO3-N and Cl- in groundwater and manure filtrate (NO3-N ≤ 150 mg L-1 and Cl- ≤ 1300 mg L-1; see Sect. 3.2). These maximum values of NO3-Ni and Cli correspond to the minimum fm. The value of fd was assumed to be the mean fd estimated from NO3- isotopes using Eq. (2), and NO3-NiCli was required to be within 1 standard deviation of the estimate from Eq. (7).

The resulting estimates of fm are reported as the mid-range, with uncertainty described by the minimum and maximum values. Larger values of fm indicate less mixing (a shorter path for advection–dispersion) and suggest a source close to the well. Smaller values of fm indicate extensive mixing (a longer path for advection–dispersion) and suggest a source further away from the well. The relative contributions of mixing and denitrification to NO3- attenuation at each site were evaluated by comparing fm and fd for each sample. This analysis was conducted using isotope values from the samples collected on 1 May 2013 at CFO1, which were combined with the Cl- and NO3-N data from 6 June 2013. At CFO4, results from stable isotopes collected on 27 October 2014 were combined with Cl- and NO3-N data collected on 7 October 2014.

The geology at CFO1 consists of clay and clay–till interspersed with sand layers of varying thickness to the maximum depth of investigation (20 m BG, bedrock not encountered). Hydraulic conductivities (K) calculated from slug tests on wells ranged from 1.2×10-7 to 4.2×10-5 m s-1 (n=10) for sand, 1.1×10-8 to 2.8×10-8 m s-1 (n=2) for clay–till, and 1.6×10-9 to 3.0×10-7 m s-1 (n=8) for clay. Depth to the water table throughout the study site ranged from 0.5 m at DMW14 to 3.8 m at DMW11. Seasonal water table variations were about 0.5 m with no obvious change in the annual average during the 6-year measurement period. Water table elevation was highest at DMW10 and DMW1 on the west side of the site and lowest at DMW11 on the northeast side of the site (see Supplement). Measured heads indicate groundwater flow from the vicinity of the EMS to the northeast and southeast. Mean horizontal hydraulic gradients at the water table ranged from 4.4×10-3 to 1.4×10-2 m m-1. Vertical gradients were predominantly downward in the upper 20 m of the profile (mean gradients ranging from 1.8×10-3 to 0.18 m m-1), with the exception of DMW11 where the vertical gradient was upward (mean gradient -2.8×10-2 m m-1). Using the geometric mean K for the sand (5.0×10-6 m s-1) and a lateral head gradient of 1.4×10-2 m m-1 yields a specific discharge (Darcy flux, q) of 2.2 m yr-1. Assuming an effective porosity of 0.3 (Rodvang et al., 1998), the average linear velocity (v‾) is 7.4 m yr-1. This suggests that, in the absence of attenuation by mixing or denitrification, agriculturally derived NO3- could have been transported through the groundwater system by advection about 400 m since 1960 and 630 m since 1930.

The geology at CFO4 consists of about 5 m of clay (with minor till) underlain by sandstone, to the maximum depth investigated (20 m BG). Hydraulic conductivities measured using slug tests on wells were 1.0×10-8 to 1.0×10-5 m s-1 (n=12) for the clay and sandstone (many shallow wells were screened across the clay–till and into the sandstone) and 1.0×10-5 to 2.9×10-5 m s-1 (n=4) for the sandstone. The depth to water table ranged from 1.0 to 3.4 m, increasing from west to east across the study site. Seasonal water table variations were on the order of 1.5 m with water table declines on the order of 0.3 m yr-1. The horizontal hydraulic gradient was consistently from west to east, with a mean gradient at the water table of 3.9×10-3 m m-1 between BC2 and BMW2 and 4.3×10-3 m m-1 between BMW2 and BMW7. Vertical hydraulic gradients were 4.2×10-2 to 4.6×10-2 m m-1 downward. Using the geometric mean K for the site (2.9×10-5 m s-1) and a lateral head gradient of 4.3×10-3 m m-1 yields a q of 0.4 m yr-1. Assuming an effective porosity of 0.3 yields a v‾ of 1.3 m yr-1. These values suggest that, in the absence of attenuation by mixing or denitrification, anthropogenic NO3- could have been transported through the groundwater systems about 10 m by advection between 1995 and the time of sampling.

Agriculturally derived NO3- was generally restricted to the upper 20 m (or less) at CFO1 (NO3-N ≤ 0.2 mg L-1 and Cl- ≤ 57 mg L-1 in seven wells screened at 20 m). The one exception was DP11-12b, which had up to 4.1 mg L-1 of NO3-N. The southeast portion of the site also does not appear to have been significantly contaminated by agriculturally derived NO3-, with NO3-N concentrations < 1 mg L-1 in five water table wells (DMW4, DMW6, DMW14, DMW15, DMW16). In DMW6, Cl- and TN concentrations were elevated (see Supplement) but NO3-N concentrations were <2 mg L-1. Collectively, these data suggest the catch basin is not a significant source of NO3- to the groundwater at this site.

Leakage of manure slurry from the EMS at CFO1 is clearly indicated by the data from DMW3, which feature the highest concentrations of TN in groundwater (up to 548 mg L-1) and elevated Cl-, HCO3-, and DOC in concentrations similar to EMS manure filtrate (see Supplement). Nevertheless, NO3-N concentrations in this well were consistently low (1.1±2.7 mg L-1, n=22). The potential for nitrification in the vicinity of this well is indicated by NO2-N production (2.7±8.3 mg L-1, n=22). However, the data demonstrate that only a small proportion of the NH3-N in DMW3 (373.4±79.4 mg L-1, n=22) could have been converted to NO3- within the subsurface (NO3-N in groundwater ≤ 66 mg L-1). Further work is required to assess the importance of cation exchange as an attenuation mechanism for direct leakage from the EMS at this site.

Contamination by agricultural NO3- that exceeds the drinking water guidelines (NO3-N > 10 mg L-1) was observed in four wells (DMW1, DMW11, DMW13, and DP10-2) and in the continuous core (DC15-23) (Fig. 3). DMW2 and DMW12 also had NO3-N concentrations that were elevated but did not exceed the drinking water guideline (≤3.7 mg L-1). Given the evidence of partial nitrification in DMW3 (and low NO3-N concentrations), the NO3-N/Cl- ratio of contamination from the EMS was assumed to be best represented by DP10-2, which is located directly down-gradient of the EMS. Data for this well indicate values of NO3-N/Cl- predominantly ranging from 0.1 to 0.3 with NO3-Ni / Cli estimated at 0.3±0.13 (Fig. 4).

The maximum NO3-N concentration in groundwater at CFO1 (66.4 mg L-1) was measured in core sample DC15–23 (clay at 2 m b.g.l., 7 m hydraulically down-gradient of DMW3). Pore water extracted from the unsaturated zone (sand) at the top of this core profile contained 865 mg L-1 of NO3-N and had a NO3-N/Cl- ratio of 1.04, consistent with the ratio of 0.95 in the core sample. Given this consistency, and that NO3-N concentrations in the well immediately up-gradient were low (DMW3), the NO3-N in this core sample was most likely introduced into the groundwater system by vertical infiltration or diffusion from above. In contrast, elevated NO3-N (up to 21.1 mg L-1) within the sand between 6 and 12 m depth in this core had NO3-N/Cl- ratios consistent with an EMS source (0.07 to 0.31). Stable isotope values in pore water from this sand layer do not indicate substantial denitrification (δ18O ≤ 5.9 ‰, δ15N ≤ 16.7 ‰), suggesting these ratios will be similar to the initial ratios at the point of entry to the groundwater system.

In DMW13 (33 m down-gradient from DP10-2) the ratio of NO3-Ni / Cli was 0.75±0.29, similar to the NO3-N/Cl- ratio in DC15-23 at 2 m (0.95), which is interpreted as reflecting a top-down source. The NO3- in DMW13 is therefore unlikely to be sourced solely from leakage from the EMS, and could be sourced from the adjacent dairy pens or a temporary manure pile that was observed adjacent to this well during core collection in 2015 (or a combination of EMS and top-down sources).

In DMW12 the NO3-Ni / Cli ratio was not inconsistent with an EMS source, but the hydraulic gradient between DMW2 and DMW12 is negligible, indicating a lack of driving force for advective transport from the EMS towards DMW12. This is also the case for well DMW1, which is up-gradient of the EMS but had elevated NO3-N concentrations (6.5±3.6, n=18). The source of nitrate in these wells is therefore unlikely to be related to leakage from the EMS, but alternative sources (i.e. nearby temporary manure piles) are not known.

Well DMW11, 470 m from the EMS, had consistently low NO3-N/Cl- ratios (<0.05), similar to DP10-2, but estimates of Cli were 3 times higher than Cli for DP10-2 (Fig. 4b). NO3-Ni and Cli estimated for DMW11 were consistent with measured values in that well, indicating a local top-down source. Well DMW11 is located hydraulically down-gradient of feedlot pens and adjacent to a solid manure storage area, in a local topographic low. Elevated NO3-N in this well is therefore interpreted to be from surface runoff and top-down infiltration, rather than lateral advection from the EMS.

At CFO4, measured data indicate that effects from agricultural operations on NO3- concentrations in groundwater are restricted to the upper 15 m of the subsurface. NO3-N concentrations in wells screened at 15 m depth were <0.5 mg L-1, with the exception of one sample from BP10-15w (May 2012) with 4.3 mg L-1 of NO3-N. Water table wells in the west and north of the study site (BC1, BC2, and BC3) also indicate negligible impacts of agricultural operations, with Cl- < 10 mg L-1 and NO3-N < 0.1 mg L-1.

Concentrations of NO3-N > 10 mg L-1 were measured in three water table wells (BMW2, BMW3, BMW4) adjacent to the EMS, indicating that they have been impacted by the EMS (Fig. 5). Of these, BMW2 had much higher Cl- concentrations (502±97 mg L-1, n=22 in BMW2 compared to 182±81 mg L-1 in BMW3 and 188±74 mg L-1 in BMW4), and therefore lower NO3-N/Cl- ratios (<0.05). Cl- concentrations in BMW2 were consistent with concentrations in the EMS suggesting direct leakage, while stable isotopes of NO3- and initial concentrations (NO3-Ni≥127 mg L-1) indicate substantial denitrification (Table 3, Fig. 6). The NO3-Ni / Cli ratio in BMW2 is consistent with measured NO3-N/Cl- in BMW4, which therefore likely reflects leakage from the EMS without denitrification (consistent with stable isotope of values of NO3-).

Given that the estimated subsurface travel distance during operations at this site is 10 m, agriculturally derived NO3- in other wells not immediately adjacent to the EMS is unlikely to be related to leakage from the EMS. Wells BMW5 and BMW7 are 60 and 140 m hydraulically down-gradient from the EMS, respectively. NO3-Ni / Cli ratios in these wells were not inconsistent with BMW2 (i.e. the range of values overlap), but given the distance from the EMS the source of NO3-N in these wells is most likely the adjacent dairy pens. Concentrations of NO3-N > 10 mg L-1 were also measured in BC4, which is located 95 m hydraulically up-gradient of the EMS. The ratio of NO3-Ni / Cli at BC4 was the highest at CFO4 (0.6) and did not overlap with BMW2. The NO3- in this well is interpreted to have been sourced from an adjacent manure pile, which was observed during the study.