Transport and Fate of Nitrate in a Saturated Buffer Zone as Assessed With a Chloride Tracer Test

Agriculture is a leading contributor of surface water pollution, contributing nutrients such as nitrogen (N), to aquatic environments (Anderson et al., 2014). N is a primary nutrient required for the growth and development of plants. The need to increase crop yield has led to the increased application of N-based fertilizers to agricultural fields (Lemke et al., 2011; Miller et al., 2011; and Robertson and Saad, 2013). Concomitant with the increased application, nitrate as nitrogen (NO₃-N) concentrations in rivers and reservoirs throughout agricultural regions have risen (Gentry et al., 1998; David and Gentry, 2000). N-fertilizers and organic nitrogen in manure are major sources of nitrogen pollution as crops do not assimilate all of the applied N (Lutz et al., 2020).

The use of N-fertilizers in the Midwest of the United States has increased since the mid-20th century. Not all of the N-fertilizer is utilized by the crops, which has led to elevated concentrations of nitrate in surface waters (Peterson et al., 2001). Excessive nitrate accumulation degrades freshwater systems and causes changes to biodiversity and the death of aquatic organisms (David and Gentry, 2000). For example, the waters within the Mississippi River Basin register some of the highest concentrations of nonpoint source nitrate (NO₃^-) in the world (Hypoxia Task Force, 2018). The algal blooms and hypoxic zones in the Gulf of Mexico and Lake Erie illustrate clear manifestation of the negative impact of NO₃^- export into water bodies. According to Keeney and Hatfield (2008), the Illinois River, a tributary of the Mississippi River, contributes from 15 to 20 percent of the total nitrogen that is transported into the Gulf of Mexico.

In the Midwest region, particularly in Illinois, the soil typically has a high water table that hinders the growth of most crops. Excessive soil moisture stunts plant growth due to an oxygen deficiency in the root zone and a reduction in nitrogen uptake resulting from denitrification and leaching (Kaur et al., 2017). To improve soil drainage and increase crop yield, farmers have installed subsurface draining systems (tiles) to drain the soil water directly into surface water bodies (Fausey et al., 1995). However, the tiles short-circuit the roles of the soil in the nitrogen cycle, reducing or eliminating the opportunity for denitrification, plant uptake, and microbial immobilization. Whereas the installation of tile drains has successfully opened additional lands for agricultural developments and increased crop yield, the short-circuiting process directly contributes to 52 percent of nitrate entering the Gulf of Mexico (David and Gentry, 2000). Removal of the tile drains to protect water bodies is not practical; thus, a solution must be found to mitigate the negative impact of tiles on water quality.

In 2008, a national strategy action plan was implemented to address hypoxia in the Northern Gulf of Mexico and to improve water quality in the Mississippi River Basin (USEPA, 2017). In 2015, Illinois developed its own plan, the Illinois Nutrient Loss Reduction Strategy, with the goal of reducing nitrogen and phosphorus loads by 45 percent in the long term, and interim goals of 15 percent reduction in nitrate and 25 percent reduction in phosphorus by 2025 (IEPA et al., 2019). Four best management practices for agricultural nonpoint sources were proposed to achieve these goals: free water surface–constructed wetlands, denitrifying bioreactors, controlled drainage, and saturated riparian buffer zones (SRBs) (Carstensen et al., 2020). With an SRB, tile-drain water is redirected into a riparian buffer zone where the water travels as groundwater to the stream. While being transported through an SRB, the nitrogen-rich waters encounter conditions favorable for denitrification, assimilation, plant uptake, and dilution of the nitrate (Jaynes and Isenhart, 2014; Carstensen et al., 2020).

In the Midwest, widely implementing SRBs could result in a 5 to 10 percent reduction of the estimated N load from land drained by tile systems (Chandrasoma et al., 2019). However, the reported efficiencies of SRBs to reduce N loading has been highly variable. After monitoring six SRBs in Iowa with 17 site-years of data, Jaynes and Isenhart (2019) reported nitrate removal between 8 and 84 percent. Groh et al. (2019), monitoring three SRBs in the Midwest, concluded that cumulative denitrification could account for up to 77.3 percent of total nitrate removed. For a period from September 2016 to February 2017, Brooks and Jaynes (2017) observed 61 percent loss of nitrate loading across seven active buffers in Iowa, Illinois, and Minnesota. SRBs have been shown to lower NO₃-N concentrations through denitrification, plant uptake, and dilution (Jaynes and Isenhart, 2014; Groh et al., 2019; Miller et al., 2019; and Bosompemaa et al., 2021).

Research concerning the use of SRBs as means to remove nitrate from agricultural waters has highlighted the variability among systems to reduce nitrate. One aspect associated with removal is the travel time of the water within the SRB, which has been reported on the order of weeks (Jaynes and Isenhart, 2019). This work reports the use of a tracer test to identify travel times of diverted tile water through an SRB and the use of the tracer test data coupled with an end-member mixing model to assess the fate of the nitrate. The results highlight the heterogeneity of the system and suggest that the loss of nitrate is spatially and temporally variable.

The study area is in central Illinois (40.614382°N, 89.023542°W) and encompasses a floodplain that was once farmed but has been restored to a grassland (Figure 1). A third order stream, Tributary 3 (T3), serves as the western boundary of the study area. Similar to other Illinois streams (Mattingly et al., 1993; Becker and Peterson, 2022), T3 has been modified and receives tile drainage from a tile system draining a field to the east. The segment of stream adjacent to the site has not been modified in the last 30 years. The agricultural field is located approximately 90 m to the east and is separated from the property by a county road.

Across the site, the surface material (0–0.6 m) is dark organic-rich topsoil, which is underlain by a clay loam (0.6–1.5 m). Between 1.5 and 2 m, the clay loam coarsens with an increasing sand and gravel percentage. The coarse-grained medium is underlain by blue-gray, dense diamicton belonging to the Tiskilwa Till member of the Wedron Formation deposited during the Wisconsin glaciation (Weedman et al., 2014). The thickness of the diamicton in the region ranges from 30 to 45 m terminating at Silurian dolomite bedrock (Wickham et al., 1988). The organic matter content in the top 60 cm is 6.0 percent but decreases to 2 percent at a depth of 90 cm (Sanks et al., 2015; Bosompemaa et al., 2021).

Surface water infiltrates through the soils to the sand and gravel zone where it then flows horizontally. The sand and gravel waters represent the “shallow” waters, whereas the waters obtained from the diamicton are referred to as “diamicton” waters. The shallow and the diamicton waters have distinct geochemical signatures. Waters in the sand and gravel zone are a bicarbonate-rich water as compared with waters derived from the diamicton, which are sulfate-rich (Akara et al., 2016). The sand and gravel zone has a hydraulic conductivity (K) value of 2.2 × 10⁻⁵ m/s, calculated from the geometric mean of slug test results from 17 wells. At a depth of 1 m, the average porosity is 0.32 (Sanks et al., 2015). The depth to groundwater is approximately 2 m during the dry season and less than 0.5 m during the rainy seasons when the tile is running. Locally, the groundwater flows from the east to west with flow toward the stream T3 (Figure 1).

The study site has been modified with the installation of an agricultural treatment system (control box) that directs the tile water (agricultural runoff) into the subsurface within the riparian area, creating an SRB. The control box consists of three chambers separated by a set of stoplogs. The stoplog between the upper and middle chambers has a 45° v-notched weir that is positioned 0.18 m below the land surface; the elevation of the weir is 227.85 m above mean sea level (m.a.m.s.l.). The upper chamber receives the agricultural runoff, whereas the middle chamber redirects the water into a 15-cm outlet pipe that bifurcates into three distribution tiles positioned 1 m below the surface. The three tiles have individual lengths of 95 m (eastern tile), 110 m (middle tile), and 50 m (western tile) (Figure 1). The distance from the stream, which has a water surface elevation of approximately 225.70 m.a.m.s.l. when the tile is running, to the closest tile varies, ranging from roughly 15 m to slightly more than 50 m in distance. The system is designed to deliver excess tile water to the stream when the tile discharge exceeds the capacity of the middle chamber; however, continuous monitoring indicates that water has not breached the middle chamber. Thus, all water entering the middle chamber is delivered to the distribution tile system. The tiles start to run from late winter (late February–early March) to late spring (late May–early June).

Within the study site, 35 observation wells have been installed; each well has a 0.75-m long screen. A series of nested wells, wells 2, 4, 6, 8, 10, and 12, are aligned with groundwater flow and intersect the distribution tiles (Figure 1). The nested wells consist of four wells, labeled A through D. The A and B wells, installed to a depth of 3.8 and 3.0 m, respectively, allow collection of water from the diamicton. Wells C and D have screened depths that end at 2.3 and 1.5 m, respectively, and are designed to collect groundwater samples from the sand and gravel zone. The other 11 wells are screened to a 2.3-m depth. The vertical hydraulic gradient calculated from the nested wells 2, 4, 6, and 8 consistently have a downward vertical gradient, whereas, near wells 10 and 12, an upwelling of waters from the diamicton into the sand and gravel zone has been observed. Water chemistry data confirm the upwelling and indicate that upwelling may also occur near well 20 (Akara et al., 2016).

A tracer test was conducted within the SRB in the spring of 2021. The tracer test started on March 9, 2021 (day 0) following sustained flow within the tile for the previous seven days. Six kilograms of sodium chloride (NaCl) were mixed with 19 L of water and added to the upper chamber of the control box. The injection into the upper chamber of the control box allowed for the incoming tile water to further mix with the tracer solution before entering into the middle compartment, where it was diverted into the SRB through the diversion tiles.

Groundwater samples were taken from the wells within the SRB and the control box prior to the start of the tracer test to ascertain the background concentration of the major anions. The first sampling post-injection occurred after 7 days on March 16, and sampling continued for 52 days following injection. Precipitation (recharge) events occurred on days 24 and 34. The recharge events impacted the breakthrough curves for the Cl^- (Figure 2). For all locations, in situ water quality parameters, including specific conductance, dissolved oxygen (DO), and temperature, were recorded using a YSI 85 probe. Water level measurements in the wells were recorded with an electronic water tape. Prior to sample collection, the wells were purged for stabilization of the in situ parameters. From each sampled well, withdrawn water was filtered through a 1-µm glass fiber filter and collected in 60-mL high-density polyethylene containers. Samples were stored on ice in the field and while in transit. Samples were stored at 4°C until being analyzed using a DIONEX ICS-1100 ion chromatograph for major anions, chloride (Cl^-), bromide (Br^-), nitrate as nitrogen (NO₃-N), and sulfate (SO₄^2-) following the U.S. Environmental Protection Agency method 300.1 (Hautman and Munch, 1997).

Post-tracer groundwater sampling started 7 days after the initiation of the tracer test. Initially, the planned sampling frequency was once a week; however, the first sampling revealed elevated Cl^- concentrations in a well, and the sample frequency increased to three times a week. Each sampling event followed the same protocol as the pretest sampling. Samples were collected from wells 4C, 4D, 6C, 6D, 8C, 8D, 10C, 10D, 12C, 12D, 19, 20, 21, and 22 during the 52-day study duration.

The concentrations calculated for the well (NO₃-N_w) represented the theoretical NO₃-N concentration assuming dilution is the only mechanism acting on the NO₃-N. However, nitrate is not conservative and can be added via nitrification and removed through denitrification or plant uptake from an SRB. Differences between the modeled NO₃-N_w concentrations, Eq. (6), and the measured NO₃-N concentrations quantify the loss or gain of nitrate from the water as it travels from the control box to the observed wells. A measured NO₃-N concentration for a given well that is greater than the modeled concentration from Eq. (6) implies that there is addition of nitrate, i.e., nitrification. A measured concentration less than the modeled concentration indicates nitrate removal, which could be due to denitrification or plant uptake.

The background concentration of the Cl^- in the groundwater and the waters of the control box were stable prior to the start of the test (Figure 2). After injection, concentrations in the waters of the downgradient wells increased, whereas the concentrations within the waters of the control box decreased. The concentrations of the Cl^- in the groundwater as measured from the wells downstream of the diversion tile ranged from as low as 2.02 mg/L to a maximum of 15.49 mg/L (Table 1). Wells 4C and 4D exhibited no elevated Cl^- concentrations, whereas wells 6C and 6D experienced higher Cl^- concentrations following the recharge events when all other wells experienced dilutions. Given the lack of a response in Cl^- concentrations, well 6 was interpreted as not receiving input from the upgradient tiles.

Recorded peak concentrations were observed on either day 33 or 37 of the test despite the recharge events on days 24 and 34 that diluted concentrations in the well waters (Figure 2 and Table 1). Velocities for waters traveling from the tiles to the wells, calculated from Eq. (1), ranged from 0.03 to 0.34 m/day with a mean groundwater velocity of 0.19 m/day (Figure 3 and Table 2). Using Eq. (2), the calculated velocity values, a horizontal gradient of 0.002, and an effective porosity of 0.032, a magnitude lower than the measured porosity (Sanks et al., 2015), the hydraulic conductivity (K) values ranged from 5.31 × 10⁻⁶ m/s to 6.32 × 10⁻⁵ m/s (Figure 3). Among the wells, no significant difference was observed between the K values obtained from the tracer test (M = 3.43 × 10⁻⁵, SD = 1.95 × 10⁻⁵) and K values calculated from the slug test (M = 2.65 × 10⁻⁵, SD = 2.02 × 10⁻⁵; t(5) = 0.741, p = 0.492). The similarity between the tracer hydraulic conductivity and slug test hydraulic conductivity values provided validation of the tracer test results.

Results from the mixing model are grouped into two sets. The first set resulted from quantifying the proportion of tile water to groundwater in the waters received in the downgradient wells (Eq. 5), and the second set was generated by calculating the expected (modeled) concentration of the NO₃-N in the well water given the mixing of waters (Eq. 6). The difference between the modeled concentration and the measured concentration in the well water indicated either an addition or a removal of NO₃-N. To calculate the proportion of tile water to groundwater, the time to peak concentration (t_p) was incorporated by using concentrations of tile water (Cl_t) and groundwater (Cl_g) t_p days prior to the day of sampling, using the concentration measured in the well (Cl_w) for the day of the sampling. To determine Cl_t for the days in which a sampling was not performed, the concentrations were linearly interpolated between the preceding and following sampled concentrations dates. The time series nature of the samples collected generally exhibit temporal continuity, so using that method of interpolation was logical. This methodology resulted in a limited set of the data because the calculation of NO₃-N concentrations could only start after the day the Cl^- arrived at a given well. Thus, results of the mixing model start the day of the observed peak (t_p).

The mixing of tile water and groundwater varied spatially and temporally during the duration of the test. Among the wells, mean ratios (Q_t/Q_w from Eq. [5]) ranged from 0.09 to 0.72 (Table 3). The variability of the ratios suggested heterogeneities within the medium had an influence on the mixing that occurs within the SRB. Modeled NO₃-N_w concentrations for the downgradient wells, Eq. (6), indicated that the SRB both contributed and removed nitrate to and from the system (Figure 4 and Table 4). Up to day 43, wells 8D, 10C, 10D, 12C, 12D, and 20 experienced measured concentrations that were two times larger than the predicted concentrations, indicating that nitrate was added to the system (Figure 4). In the four other wells, measured NO₃-N concentrations were below the modeled concentrations. By day 47, waters in all the wells recorded NO₃-N concentrations below the modeled concentrations, signifying a loss of nitrate in the system.

The DO concentrations in the downgradient wells ranged from 0.88 to 11.21 mg/L (Figure 5). Pre-test and early DO concentrations were generally the highest observed concentrations. Concentrations began to decrease following day 7. At day 36, only wells 10C and 19 had DO concentrations higher than 4.5 mg/L, and by day 52, all wells had concentrations less than 4.5 mg/L.

During the tracer test, all water entering the control structure was redirected into the diversion tiles. The two-component mixing model verified that the downgradient wells received tile waters that had mixed with the incoming groundwaters. The incorporation of the tile water into the groundwater was not uniform throughout the system as noted by the spatially variable ratio of tile water to groundwater calculated for the wells. The variability could be a result of either heterogeneity of the glacial materials or a function of the error associated with the interpolation of Cl^- concentrations required to ensure the correct lag time. There is the limitation of using a two-component mixing model when the frequency of sampling is not optimal (Anderson et al., 2014), but we feel the observed nitrate losses provided viable results, which aligned with other studies (Utt et al., 2015; Jaynes and Isenhart, 2019).

The mixing model results identified periods of nitrate addition and removal by the SRB. The nitrate removal ranged from 23 to 97 percent. Whereas this is a large range, the values are consistent with other reported values. Utt et al. (2015) reported that nitrate reductions range from 23 ± 28 percent in the SRBs within the Midwest. In the fall fallow season, Brooks and Jaynes (2017) documented a 61 percent reduction. Jaynes and Isenhart (2019) provided the most comprehensive saturated buffer assessment to date with monitoring of six saturated buffers in Iowa. They reported annual nitrate loss ranging from 7 to 92 percent in individual years and averaged 44 ± 26 percent (median, 35 percent) across all site years. In a review paper, Johnson et al. (2023) summarized from five studies with proper designs total N load reductions ranging from 7 to 92 percent.

With all tile water diverted into the system, the SRB had the potential to remove a significant quantity of the nitrate transported within the agricultural discharge. Nitrate removal could occur via denitrification of nitrate to N₂ or N₂O gas, plant uptake, microbial assimilation, and dissimilatory nitrate reduction to ammonium (Lutz et al., 2020). These processes take time to occur, and the time required for tile waters to move through the SRB is an important aspect for the utilization of SRBs as a nutrient-reduction strategy. While quantifying the effectiveness of an SRB to reduce tile NO₃-N concentrations in eastern Iowa, Streeter and Schilling (2021) reported denitrification began to occur within less than a day. Within a span of 7 days, they reported a 10-fold reduction from 15 mg/L NO₃-N to 1.5 mg/L. Applying their results to this work, the observed travel times of 33 to 37 days for waters from the diversion tiles to the wells should have provided sufficient time for removal of nitrate within the T3 SRB.

The transport and behavior of nitrate reflects the conditions within the SRB. As the tile began to flow and saturate the vadose zone, the DO concentrations within the waters reflected aerobic conditions conducive to nitrification. Within the studied SRB, previous works observed increased concentrations of NO₃-N in water during the non-growing season (Miller et al., 2019; Bosompemaa et al., 2021). Bosompemaa et al. (2021) hypothesized that the degradation of the plant material during the non-growing season generated nitrate in the vadose zone. During multiple 24-hour sampling events in the winter and spring, Miller et al. (2019) observed rising NO₃-N concentrations of the groundwater, attributing the increase to nitrification associated with the assimilation of decaying plant material. The timing of the observations by Miller et al. (2019) represented the same time period as the initiation of the tracer test. Whereas the tile diversion system had been running for a couple of days prior to the start of the test, the unsaturated zone of the SRB was beginning to become saturated. We posited that the nitrate accumulated from the decay and degradation of plant material within the unsaturated zone during the fall and winter was being flushed from the unsaturated zone, contributing to the elevated (above modeled) NO₃-N concentrations. The NO₃-N concentrations above the modeled concentrations were measured in wells with DO concentrations above 4.5 mg/L, the threshold observed for denitrification by Anderson et al. (2014). As the soils of the SRB saturated, DO concentrations decreased, falling below 4.5 mg/L in all well water by the conclusion of the test (Figure 5). The lower DO concentrations coincided with measured NO₃-N concentrations falling below the modeled concentrations, which indicated a loss or removal of nitrate from the system.

Plant uptake and denitrification have been postulated as means to remove nitrate within the T3 SRB (Miller et al., 2019; Bosompemaa et al., 2021). Denitrification leads to the permanent removal of nitrate, whereas the plant uptake is a temporal removal of nitrate. The occurrence of denitrification requires anaerobic conditions and labile organic carbon. When the agricultural runoff enters the SRB, the runoff displaces oxygen in the soil, creating anaerobic conditions. The average measured dissolved oxygen during the period of the tracer test is 4 mg/L. Streeter and Schilling (2021) observed that denitrification is mostly dominant when the dissolved oxygen rate within an SRB ranges from 2 to 4.5 mg/L. With DO concentrations below 4.5 in the T3 SRB, the study site had the required anaerobic conditions required for denitrification to take place.

A soil organic carbon content greater than 0.75 percent is recommended to induce denitrification (USDA-NRCS, 2018; Jaynes and Isenhart 2019). The surficial materials, the top 1.5 m, of the T3 SRB comprise silts and clays, rich with organic matter. Bosompemaa et al. (2021) reported an average organic matter content at the study site to be 6.0 percent at a depth of 60 cm, indicating that there is sufficient organic matter for denitrification to occur within the study site.

Whereas plant uptake has been proposed as a contributing factor to the observed loss of nitrate within the site, the uptake requires the plants to be growing. As a restored prairie, the study site is dominated by switch grass, which can recover 66 percent of applied nitrogen (Bransby et al., 1998). However, at the initiation of the tracer test, the plants were still dormant. In April, the plants were beginning to emerge from dormancy, and growth was observed. Whereas the removal of nitrate aligns with the reemergence of the plants, an observation consistent with other studies (Munoz et al., 1993; Clément et al., 2003; and Li et al., 2016), the observed reduction of nitrate in wells prior to plant growth supported denitrification as the main removal mechanism (Miller et al., 2019; Lutz et al., 2020). Additional support for denitrification comes from Groh et al. (2019); they reported in situ denitrification within SRBs via the acetylene inhibition method and found out that denitrification accounted for between 4 and 77 percent of the total nitrate removed within the SRB.

An expectation would be that waters sampled from wells farther away from diversion tiles would have experienced greater nitrate removal given the extended travel distance and time. However, the wells closer to the diversion tiles experienced greater loss of nitrate. Wells 19 and 22 experienced average removal of nitrate representing 91 and 59 percent of the nitrate delivered from the tile waters, whereas wells farther downgradient, wells 10 and 12, witnessed mean differences representing an addition of nitrate. However, by the end of the test, NO₃-N concentrations for the waters in all wells show a loss of nitrate with maximum loss ranging from 23 to 97 percent. Additionally, similar research conducted by Anderson et al. (2014) showed that there was no strong relationship between the amount of nitrate removed and distance away from diversion tiles. They observed some wells closer to diversion tiles that had higher nitrate removal, and their results compare quite well with the results from our test.

The results from the tracer test highlighted periods of different nitrate behavior, addition and removal, within the SRB. When the SRB was being saturated at the beginning of the tracer test, the introduced tile waters flushed the vadose of nitrate accumulated from the breakdown of plants. Limited denitrification occurred as the presence of aerobic conditions near wells would prevent denitrification. As the SRB continued to receive tile water, aerobic conditions transitioned toward anerobic conditions, which corresponded to a coinciding decrease in NO₃-N concentrations that were below the expected NO₃-N concentrations. As time progressed and the plants became more active, greater loss of nitrate was calculated. Overall, the SRB witnessed nitrate reductions between 23 and 97 percent. The observed reductions compared favorably with other published studies exploring SRB and reinforced the effectiveness of SRBs to remove nitrate. Thus, SRBs remain an effective nitrate loss-reduction practice, However, SRBs alone will likely not be sufficient to meet the water quality goals set out by the Illinois Nutrient Loss Reduction Strategy

The authors thank three anonymous reviewers for their comments and feedback, which have improved the paper. This research was funded by the Illinois Water Resources Center, grant number 079901-17840. The authors thank the City of Bloomington (IL) for access to the study location. All authors contributed to the study conception and design. Material preparation and data collection by A.S. and E.W.P. Data analysis and visualizations by A.S. and E.W.P. The first draft of the manuscript was written by A.S. with A.S. and E.W.P. working on the final paper. All authors read and approved the final manuscript.