Fate and consequence of nutrients at an abandoned feedlot, Glacial Ridge National Wildlife Refuge, Minnesota, USA

Excessive agricultural nutrients – nitrogen (N) and phosphorus (P) – in soils and waterways are a global problem (Carpenter et al. 1998; Smith 2003), with abundant quantities generated in feedlots. In the U.S. alone, cattle feedlots generate an estimated 18 million metric tons of nutrient-rich manure per year (Sweeten 2000), which affects runoff water quality. On average, an adult cow excretes 55 kg N and 18 kg P per year (Goolsby et al. 1999), although these nutrient amounts can vary significantly (e.g., Davis et al. 2002). Thus, for even small feedlots, the nutrients generated can easily total in the tens of thousands of kilograms. While much N is lost to the atmosphere as ammonia (Hristov et al. 2011) and some of the N and P is spread and broadcast beyond the confines of the feedlot (Everett and Vickery 2005), a large portion may be transported vertically and then laterally after leaching into soils that underlie the facility (García et al. 2012; Vaillant et al. 2009; Chang and Entz 1996). In this report, we suggest that feedlots may create “hot spots” of N and P that can serve as a potentially long-term source of nutrients, which can lead to downstream eutrophication, harm nearby wetlands, and degrade natural plant communities.

The fate of nutrients in feedlots is controlled by biogeochemical processes that depend significantly on hydrogeological factors. Nitrate from agricultural sources, including manure and feedlots, has resulted in the widespread degradation of water resources (i.e., Power and Schepers 1989), with limnic and marine zones effected by runoff from agricultural watersheds, such as the Gulf of Mexico near the outlet of the Mississippi River (i.e., Rabalais et al. 2002). In response to this major environmental problem, much research to-date has focused on the transport and transformation of nitrate that occurs by leaching and denitrification (i.e., Korom 1992; Schipper et al. 2010). Dual analyses of oxygen and nitrogen isotopes in nitrate (reviewed by Kendall 1998), in particular, reveal the source and fate of nitrate in a variety of environments, including those that host livestock (Fenech et al. 2012). In many cases, denitrification processes can be tracked by the dual isotope approach (Böttcher et al. 1990; Fukada et al. 2003).

In contrast to nitrate, phosphorus is typically the limiting nutrient in freshwater eutrophication (Schindler 1981; Vollenweider and Kerekes 1982). P is not lost to the atmosphere and is instead sequestered by precipitation of mineral phases and adsorption onto Fe-hydroxides, clays, carbonates, and organic matter. In some circumstances, P can be mobilized and lost to nearby streams and wetlands if the soils hold excessive P concentrations, contain abundant organic matter that may cause chemical reduction, or are deep and permeable (Eghball et al. 1996; Sims et al. 1998; Eghball 2003). Recent research indicates subsurface P transport is not negligible (King et al. 2015). Furthermore, large volumes of solid manure may lead to saturation of P retention capacity, which will enhance the capacity for P transport (Ciapparelli et al. 2016).

The fate of P in soil is determined by a variety of competing processes. Depending on the pH and oxidation reduction conditions, P can be precipitated as Ca-, Fe-, or Al-phases, or can be adsorbed onto organic or clay particles (Djodjic 2001; Sekhon 2002). Phosphorus adsorption, however, competes with fixation of organic acids generated from the manure (Chardon and Schoumans 2002) and the loss of adsorption capacity by the formation of Fe and Al complexes (Bolster and Sistani 2009). Once soil adsorption capacity is reached, which may occur in soils influenced strongly by organic matter (Ciapparelli et al. 2016), P can be mobilized (Campbell and Racz 1975; Eghball 2003; Glaesner et al. 2012) and transported to the water table where low oxygen and concomitant reduction may increase P solubility. Thus, coupled with the subsurface movement of organic matter, the geological characteristics of soils and near-surface sediments – texture, composition, and the oxidation-reduction conditions in the subsurface – likely control the transport and concentration of P in and around feedlots.

Coupled with excessive input of nutrients, site operations adjacent to pens and large numbers of stock lead to disturbance of native vegetation communities and establishment of invasive exotic species. Tognetti and Chaneton (2015) showed how mowing in pastures (and often done adjacent to feedlots) combined with N addition led to rapid invasion by short-lived exotic forbs and later by exotic perennial grasses. Also sensitive to disturbance are wetlands and riparian zones that may lie near or adjacent to disturbed areas, including feedlots. For example, alien and invasive species such as Phragmites australis and Lythrum salicaria are clear indicators of wetland disturbance along Connecticut (USA) roadways (Moore et al. 1999). In Illinois (USA) restored wetlands, non-native invasion and dominance were found to be controlled in larger measure by abiotic factors, such as nitrogen availability and disturbance, rather than propagule strength and resiliency, which enhanced non-native richness (Matthews et al. 2009). In other wetlands and riparian areas, disturbance combined with increasing nutrients favor invasion of monospecific, fast-growing, robust, and morphologically plastic species such as Phalaris (Kercher et al. 2007) and Phragmites (Minchinton and Bertness 2003).

Lastly, feedlot status-of-use (Vaillant et al. 2009) and age of the site (Zhu et al. 2004) are important considerations related to possible environmental degradation and contamination, with older and abandoned sites posing a greater risk because of diminished management and typically fewer environmental protection regulations that were active during operation. Mitigating “legacy” P derived from agricultural sources presents challenges because the traditional conservation approach, reducing soil erosion and transport to surface water, may not provide a sufficiently robust management strategy (Kleinman et al. 2011). Under optimum hydrological conditions, even low concentrations of legacy P in cropped fields, which are likely to have lower concentrations of P than feedlots, can contribute significant loads of P to runoff. Tillage management, field amendments, and engineered runoff filters can be used to mitigate P release and transport, but these alternatives often demand costly, focused efforts that are rarely used in agricultural operations (Kleinman et al. 2011).

In our report, we show to what degree a single, abandoned feedlot influenced and continues to affect its surrounding environment. For this case study, we investigate the spatial and temporal distribution of soil and groundwater N and P in and near an abandoned feedlot in north-central USA, which operated between 1970 and 1999. First, we describe the landscape and current distribution of nutrient concentrations at the former feedlot site. Next, in the context of soil composition and potential biogeochemical processes, we suggest reasons for the observed variability of nutrients across the site and evaluate the degree of transport into adjacent wetlands and surface-water channels. Finally, our report describes more generally the likely biogeochemical processes that determine the mobility of feedlot nutrients, along with the ecological consequences.

Site description of the abandoned cattle feedlot. Our case example, the former Crookston Cattle Company feedlot (47°43.6’ N, 96°19.2’ W), lies within the US Fish and Wildlife Service’s Glacial Ridge National Wildlife Refuge near Crookston, northwestern Minnesota, USA. (Figure 1, insert). The feedlot contained as many as 2,500 cattle at one time during its 30-year operation, although the number of livestock was generally much less, with an average of 500 head for the period of operation, including the times when the livestock was on the range. Similar numbers of sheep occupied the lot during the last few years of the feedlot’s operation (1995–2000). The pens on the northeast end of the feedlot were used more than those toward the southwest (Figure 1) (Terrill Bradford, University of Minnesota-Crookston, personal communication).

The feedlot was established on a gentle, low-relief sandy ridge. Formed roughly 9,900 years ago, the ridge developed along a prominent strandline on the eastern margin of glacial Lake Agassiz (Wright, 1972). This location was selected for development of the livestock facility because of enhanced air circulation on the ridge and ample natural subsurface drainage. The former feedlot pens and wastewater ponds cover about 15 ha and are drained by a ditch system, which conveys surface runoff into the Red Lake River, a tributary to the Red River of the North and downstream Lake Winnipeg. In recent decades, Lake Winnipeg has become increasingly eutrophic, due largely to agricultural activity and the regional expansion and increasing efficiency of drainage within the Red River portion of its watershed (e.g., McCullough et al. 2012; Wassenaar and Rao 2012).

The Natural Resources Conservation Service (NRCS) Digital Soil Survey Geographic Database (SSURGO) digital maps show that the soil series underlying the site consist of excessively well-drained Sandberg – Radium complex loamy sand along the highest axis of the ridge, Radium loamy sand or Syrene sandy loam along the flanks, and Strathcona fine sandy loam in adjacent wetlands. These soils are Mollisols with a combined O-A-B horizon that ranges from about 0.4 m in wetlands to 0.8 m in more elevated sites (NRCS 2015).

Historical imagery (Figure 2) shows that partial drainage of wetlands in the vicinity of the feedlot occurred before 1939, with legal records suggesting excavation of the main ditch to the east and south in the 1910s. Because of the sandy, excessively well-drained soils that underlie the feedlot, this area was never cultivated but used for grazing and forage before the feedlot was established in the 1970s. In contrast, effective drainage of the former wetland basins toward the east and south allowed for cultivation, which began in the 1990s. As part of the Glacial Ridge Project (Gerla et al. 2012), many of these ditches were filled in 2005 and formerly drained areas, where accessible, were planted in native perennial grasses. Neither the National Wetlands Inventory (US Fish and Wildlife Service 2015) nor SSURGO hydric soils (NRCS 2015) clearly delineate wetlands near the feedlot. In addition, only components of some of the soil series to the northeast and south are hydric; not the entire area shown actually functions as wetland. Thus, wetland-upland boundaries shown near the feedlot are generalized (Figure 2).

The wetland basin north of the feedlot was never entirely drained (Figure 2) but received feedlot runoff and waste that flowed through an unlined sewage lagoon excavated in the 1970s. Prior to that, the wetland was likely used for pasture and grazing (Figure 2, upper images), followed by no management and non-use while the feedlot was in operation, and then finally outlets were filled at least partially during the Glacial Ridge Project. This attempt at limited restoration led to a recent condition where open water occurs seasonally (Figure 2, lower right). Because of the wetter than long-term average conditions that have affected the region since 1993 (Winter and Rosenberry 1998), the presence of open water in the basin unlikely existed except recently, at least during historical times.

Groundwater and soil investigation. Groundwater monitoring began in 2002, with extensive sampling and analysis of groundwater and soil carried out in 2013 using methods described in detail by Gbolo (2014). After laying out an irregular grid with transects oriented perpendicular to the trend of the beach ridge (Figure 1 and Table S1), soil samples were collected at 63 sites, prepared, and analyzed for texture, pH, electrical conductivity, total nitrogen, nitrate, ammonium, available phosphorus (Olsen et al. 1954), total carbon, organic carbon, calcium, and iron. Samples were collected from both the combined O- and A-horizon and soil B-horizon (Table S2 and S3), so depths varied depending on the landscape position and soil series. Plant associations were mapped by combining field locations of boundaries determined with GPS together with visual identification of vegetation communities on recent Farm Service Agency aerial imagery.

Groundwater was sampled from 16 shallow water table wells at the site (Figure 1 and Table S4), which ranged in depth from 2.4–3.2 meters and had either 0.3 or 0.6 m sand-packed, mill-slot PVC, 5 cm (2 inch) diameter screens. The wells were thoroughly purged before sampling, and samples were transported on ice to the University of North Dakota Environmental Analytical Research Laboratory (UND EARL), with analyses completed within the time limit recommended by the US Environmental Protection Agency (1983). Groundwater samples were obtained from these wells with screens that were placed either across or immediately below the mean level of the water table, always within 0 to 1 m of the tension saturated zone. An additional 21 more single and nested wells, along with four open pits fitted with a standpipe were used to monitor water levels estimate lateral and vertical hydraulic gradients at the site (Gbolo 2014). Lastly, monitoring data from one monitoring well installed by the U.S. Geological Survey (G08-R 149N44W17ABAD 0000620668) (Figure 1, open circle) supplemented groundwater geochemistry presented in this report.

Surface water investigation. Surface water was sampled within wetlands adjacent to the feedlot on the northwest and southeast, and along Polk County Judicial Ditch 66, the nearest surface-water channel, which lies roughly 500 m east of the feedlot (Figure 3). Much of the water that is conveyed by the ditch is groundwater discharged from a large sand and gravel production pit that lies 800 m east of the feedlot, maintaining at least 0.006 cubic meters per second during the driest times and as much as 0.2 cubic meters per second during spring runoff. Three sites were sampled at various times between 2007 and 2013: near the outlet of the aggregate production pit, downstream about 800 m (nearest the feedlot), and roughly 1,700 m farther downstream. Along with the feedlot and adjacent wetlands, this ditch also receives runoff from surrounding former cropland, which is now reconstructed prairie and wetland. The surface-water analytes and laboratory techniques used were the same as those for groundwater analysis.

Chemical analysis. Chemical analyses of the soil extracts and water samples were performed at the UND EARL. A Shimadzu TOC analyzer (Model VCSH) was used to analyze for total carbon, inorganic carbon, and total organic carbon. The filtered (0.45 μm) unpreserved samples were used for nitrate-nitrogen, nitrite-nitrogen, and soluble or dissolved reactive phosphorus analyses. Nitrate-nitrogen and nitrite-nitrogen were measured using ion chromatography (Dionex DX-120). The filtered (0.45 μm) and acid-preserved samples were used for ammonium-nitrogen analysis. Soluble reactive phosphorus, total phosphorus, and ammonium-nitrogen were measured using a HACH DR/2010 spectrophotometer. Soluble reactive phosphorus (SRP), total phosphorus (TP), and ammonium-nitrogen (NH4-N) analyses used ammonia nesslerization, acid-hydrolysable, and ascorbic acid methods, respectively. Calcium and iron were analyzed using the EARL’s inductively coupled argon plasma-atomic emission spectrometer.

Nitrogen and oxygen stable isotopes in groundwater nitrate were collected between 2009 and 2013 and analyzed by mass spectrophotometry at the University of Nebraska – Lincoln’s Water Sciences Center using the laboratory’s VG Optima dual inlet isotope ratio mass spectrometer. Water samples for isotope analysis were collected from wells using a disposable bailer, filtered (0.45 µm), frozen at < –18°C, and shipped by overnight delivery to the laboratory in a well-insulated package.

Statistical analysis and mapping. A square grid (x = y = 11.6 m) was created from the scattered soil and groundwater analytical data using a multi-quadric radial basis function algorithm (Carlson and Foley 1991a), with an R-squared smoothing factor (Carlson and Foley 1991b) selected to highlight the trends in analyte concentration across the feedlot and its vicinity. To create a grid, the radial basis function approach uses a summation of quadric surface regression equations with coefficients determined by values of known points across the mapped area. This method generates a grid that filters local irregularities while maintaining fully the trend of the variable (e.g., Hardy 1971). The output from the gridding process was also used to estimate the volume of soil and groundwater that remains affected by feedlot-related nutrients. This was done by applying the extended Simpson’s 3/8 Rule, whereby parabolas are used to approximate z-values for each part of the curving surface. Volumes were then estimated by integrating z between the underlying clay till and the ground surface (for soil) or water table (for groundwater).

Groundwater flow. Monitoring from 2010–2014 showed that the water-table level beneath the feedlot varied within a range of 0.5 to 1 m, the largest difference occurring in the coarsest and thickest sediments, and diminishing towards the wetlands. Clay till underlies the site at greater depth, forming an essentially impervious lower boundary at a depth ranging from 0.5 m along the margins to 9 m along the northeast-southwest trending axis of the sandy ridge (Gbolo and Gerla 2013). These observations revealed that a broad, low-relief groundwater mound or crest, maintained by shallow groundwater recharge, occurs within the glacial Lake Agassiz beach sediments that underlie the feedlot.

Groundwater flow, based on water-level monitoring (Gbolo 2014), follows a simple pattern of recharge along the northeast to southwest-trending sandy ridge, accompanied by lateral movement and discharge toward wetlands on the northwest and southeast (Figure 3). This relatively uncomplicated and constrained flow system, underlain by clay till, provided conditions conducive to assess nutrient transport.

Ecohydrological setting. Plant community maps (Figure 3) show that the feedlot is flanked on the northwest and southeast by partially drained wetlands characterized by generally monospecific stands of reed canary grass (Phalaris arundinacea), hybrid cattail (Typha sp.), and common reed (Phragmites australis), with more upland sites covered by grassland vegetation dominated by a largely non-native bluegrass (Poa sp.), quackgrass (Elymus repens), field thistle (Cirsium discolor), and smooth brome (Bromus inermis), with some original (not restored) big blue stem (Andropogon gerardii) in less disturbed areas, including the areas that underlie the former feedlot animal pens.

These generally invasive plant communities tend to correspond closely to the distribution of soil series, with Typha, Phalaris, and Phragmites occurring in areas underlain by Strathcona fine-sandy loam and the well-drained Radium and Syrene series hosting Poa, Elymus, Cirsium, and Bromus. Regionally, less disturbed well-drained soils have somewhat lower concentrations of organic matter and nutrients when compared to soils on beach-ridge flanks and wetland margins, but the pattern is reversed at the former feedlot. For example, mean nitrogen and phosphorus content of the sandy soils on the ridge is at least double or as much as an order of magnitude more, respectively, even with only small differences in the organic matter content (Tables 1 and 2).

Soil and groundwater nitrogen. Soil nitrate in the shallow O- and A-horizon soils (Table 1 and Table S2) has its greatest concentrations along the margins of the feedlot pens (in excess of 10 mg/kg), with the largest concentrations associated with the waste lagoon southeast of the site (Figure 4). In contrast, soil nitrogen in the B-horizon (Table 2 and Table S3) is more strongly concentrated along the southeastern margin of the feedlot. Nitrogen distribution in B-horizon soils is similarly distributed spatially to shallow groundwater nitrate concentrations and situated almost directly below the feedlot pens, suggesting that B-horizon soil and nitrate have a similar geochemical control on concentrations. Nitrite and ammonium concentrations in water samples were all near or below the detection limit (Table S4).

Nitrogen isotopes. A large range in oxygen and nitrogen isotopes in nitrate was recorded for the feedlot and surrounding crop land, ranging from 0 to +44‰ for δ¹⁵N and 0 to +53‰ for δ¹⁸O (Figure 5 and Table S5). In comparison to nitrate, only minor concentrations of ammonium-N and nitrite-N were detected in soils and groundwater at the feedlot site, with mean concentrations for both roughly one order of magnitude less than nitrate-N (Table S4).

Soil and groundwater phosphorus. As expected in the coarse textured, aerobic soils along the beach ridge, soil and groundwater phosphorus concentrations are largest directly below the pens at the feedlot: 112 mg/kg and <0.1 – 0.35 mg/L, respectively. Results suggest minimal transport of P from the feedlot (Figure 6) when compared to the background concentrations, which are soil P at roughly 3 mg/kg and total dissolved P < 0.1 mg/L, inferred from values near the fringe of the feedlot and ungrazed/undisturbed areas elsewhere within the Glacial Ridge National Wildlife Refuge (Gbolo 2014). Note that the two “hotspots” are in roughly the same position as elevated nitrate in the soil B-horizon (Figure 4). The pattern and composition of P in the O-A soil horizon (Table S2) is similar to the B-horizon (Table S3) and therefore not shown or described in detail.

Calcium, iron, and organic carbon concentrations in soil. As noted in the introduction, the transport and sequestration of P and nitrate are likely controlled by calcium, iron, and organic carbon. These were analyzed in the soils sampled at the feedlot (Tables S2 and S3). Results showed that calcium ranges from about 3,500 to 10,000 mg/kg, with a spatial distribution that is strongly related to the sandy beach ridge that underlies the former feedlot (Figure 7). Greatest concentrations lie along the lower flanks of the ridge, and appear largest in the proximity of the feedlot lagoons.

In contrast, the largest concentrations of Fe (about 65 mg/kg) generally coincide with the southeastern flank of the ridge, and fall to less than 20 mg/kg in the wetlands that lie both to the northwest and southeast (Figure 7). Organic carbon content in the A horizon revealed a pattern that is similar to calcium (Figure 8), with the lowest soil concentration (<5%) occurring beneath the feedlot (within the beach ridge) and increasing to as much as 30% toward wetlands, and averaging 15%.

Nitrate and phosphorus in surface water. Nitrate and phosphorus concentration in surface water drained from the vicinity of the feedlot varied widely during the sampling and analysis from July 2007 through June 2013. This was also the time during which the surface water drain was retrofitted to a more natural pattern (Gerla et al. 2012), with excavation and re-grading contemporaneous with wetland restoration downstream in June 2010.

Nitrate concentration in surface waters remained well below the U.S. EPA drinking water standard of <10 mg/L nitrate-N for the upstream and downstream stations for all sample events (Figure 9). Generally, the nitrate varied between not detected (<0.1 mg/L) and 1 mg/L, with no obvious difference between the two stations, although a concentration of 2.2 mg/L was reported at the upstream – aggregate pit outlet sample site on one occasion (Figure 9).

Total phosphorus showed generally very low concentrations throughout the monitoring period 2007–2011 (Figure 9). Concentrations had greater variability at the upstream station in when compared to the lower sample location. All but four of the analyses revealed concentrations at or below 0.1 mg/L, with the other four ranging up to 0.23 mg/L at the upper sampling point in July 2007 (Figure 9). A surface-water concentration of 0.1 mg/L total P (100 ppb) has been adopted by many states as the water quality threshold.

An increasing trend in total P concentration occurred at the downstream sample site but not at the upstream monitoring station, with this latter site showing a wide range of concentration. Although not investigated further, this sharply varying concentration may correspond to periods of aggregate extraction or the presence and absence of large flocks of migratory birds in flooded aggregate pits. In contrast, the apparent rising trend in phosphorus concentration at the downstream site from 2009–2011 may have resulted from soil and channel disturbance during the ditch retrofit and then later with the restoration of a large wetland basin immediately upstream from the sample station.

Land use changes on plant communities. Aerial imagery from 1939, 1954, 2003, and 2013 shows significant changes in land use and land cover between 1954 and 2003. Prior to at least 1954, the upland areas at the feedlot were used for forage (Figure 3). Changes of ownership of the tract in the 1970s and 1980s began the conversion of pasture and forage to crops (Gerla et al. 2012). After abandonment of the feedlot, especially in the 1990s, areas to the southeast and northeast adjacent to the feedlot were drained, tilled, and cultivated with wheat, soybean, corn rotation. During the Glacial Ridge project, ditches were partially filled and the formerly drained areas were seeded with native perennial grass species, starting in 2005. Previous wetland areas that were not drained and the area encompassing the feedlot pens, however, were not reclaimed. Currently, most if not all of these areas are characterized by plant communities with dominantly non-native, generally invasive species (Figure 2). In particular, the ubiquitous hybrid Typha and Phalaris arundinacea stands in and around wetlands indicate that restoration to native species with moderate diversity will be difficult and likely impossible (Zedler and Kercher 2005), even with a return to former hydrological patterns. Diminishing nitrate and overall sequestration of phosphorus will likely do little to mitigate the disturbance related to the feedlot, with present conditions providing a niche for invasive species (e.g., James et al. 2010; Levine et al. 2003).

Oxygen and nitrogen isotopes in groundwater nitrate. Stable isotopes of nitrogen and oxygen in groundwater nitrate suggest a variety of nitrate sources at the site (Figure 5), especially manure together with nitrate- and ammonium-based fertilizers. Although it is unlikely that nitrogen fertilizers were ever directly applied to soil at the feedlot, cultivation and fertilizing equipment was stored at the northeastern end of the site. In addition, handling and storage of forage used for cattle may have leached naturally and released nutrients into soils. Fractionation revealed by the unusually large range of δ¹⁵N from about 0‰ to more than +40‰, suggests strongly that past and on-going active denitrification occurs at the feedlot (e.g., Kendall 1998; Kendall et al. 2007; Komor and Anderson 1993). Although data are not shown temporally, the largest δ¹⁵N ratios (>+30‰) were recorded for oldest samples collected (2010), and have isotopic signatures that are significantly different from nearby crop land. Three conditions must be met for denitrification to occur: 1) the presence of denitrifying bacteria, 2) sufficiently reduced, anaerobic conditions, and 3) availability of electron donors (Firestone 1982; Korom 1992). These conditions are likely met at the feedlot, with the flux of organic carbon through the groundwater flow system serving to create sufficiently reducing conditions and providing an electron donor for denitrification reactions.

The US Geological Survey has been monitoring groundwater conditions since 2003 (Cowdery et al. 2008) at a water-table well located immediately northeast of the feedlot. Data show a marked change in nitrate concentration (Figure 10), with a major drop occurring in 2005–2006, perhaps triggered by unusually dry conditions in 2006. Nitrate near the water table would have been transported to deeper levels where anaerobic conditions prevail during normal conditions; this translocation of nitrate followed by recharge and a return to normal levels of the water table may have accelerated denitrification and loss of nitrate from the groundwater system. Alternatively, recent work indicates that large changes in nitrate loss to streams can be triggered by a strong change in prevailing precipitation, or “weather whiplash” (Loecke et al. 2017). Dry springs following wet autumns in the upper Mississippi River basin have been shown to decrease nitrate concentrations in streams and rivers. Conversely, a wet spring following a dry autumn results in elevated spring nitrate concentrations (Loecke et al. 2017). Although unusually dry conditions during 2006 correlate with the sudden drop in groundwater nitrate at the feedlot, concentrations did not recover even partially when normal weather conditions and water-table levels returned (Figure 10), nor is there evidence for transport of nitrate to nearby drains (Figure 9). When considered together with stable isotope trends (Figure 5), denitrification seems to provide a better explanation for the loss of nitrate.

Phosphorus sequestration. Phosphorus tends to be sequestered by oxidizing conditions coupled with (1) formation of ferric iron and calcium-phosphate minerals, (2) adsorption onto clay, organic matter, and iron hydroxide minerals, as previously noted, and (3) assimilated and recycled by vegetation (Kaila 1949; Cole et al. 1977; Damon et al. 2014). Although the greatest potential for sequestration lies at the margins of the beach ridge and feedlot, much P remains beneath the feedlot (Figure 6). This suggests effective isolation of P by calcium, iron, and possibly wetland organic matter. These processes are likely to occur within the soil B-horizon, characterized by an accumulation of soluble or suspended organic material, clay, iron, and aluminum transported from shallower soil horizons (Table 2). The water table tends to rise and fall through this soil horizon at the feedlot, resulting in local sorption and desorption of soil P.

Elevated concentrations of calcium occur in wetlands directly adjacent to the sandy ridge (Figure 7, top). The beach ridge sediments are reworked from the coarse fraction of the Red Lake Falls Formation, which directly underlies the feedlot and contains a large proportion of carbonate rock fragments (Harris et al. 1974). One explanation for the calcium-concentration distribution may be that groundwater recharge becomes increasingly calcareous as infiltration occurs in the carbonate-rich sediments of the ridge. As this carbonate-saturated/supersaturated groundwater is discharged in adjacent wetland margins, carbonate precipitates as carbon dioxide is lost from the water (Stumm and Morgan 1996) thereby increasing the concentration of calcium in these wetland-fringe soils.

Total extractable iron shows a peculiar pattern of elevated concentration in the soil B-horizon (Figure 7, bottom) that is enriched along the southeastern margin of the feedlot and continues at elevated concentrations up to 60 mg/kg toward the southwest. One explanation may be the occurrence of heavy sand placers along the back-beach shore of glacial Lake Agassiz. Beach wave action can concentrate relatively inert minerals with high specific gravity, commonly magnetite and hematite. These heavy sand placers are well known and have been mapped along a wide range of shore line environments (e.g., Stanaway 2012; Komar 2007), although in the glacial Lake Agassiz region placers are likely small and of little value. For example, at several locations along the east coast of Australia, heavy sands accumulated in a “back-barrier wash-over facies, at the rear of swash-aligned barriers in coastal embayments” (Roy 1999), which describes features similar to the glacial Lake Agassiz-age beach sediments at the feedlot. These small iron-rich zones were likely deposited as fine magnetite and hematite sand retained in the back-barrier area during storms on glacial Lake Agassiz, later weathering to form an iron-rich zone in the soil B-horizon.

Large concentrations of both iron and calcium, which developed naturally along the margins of the beach ridge (Figure 7), suggest that P from the feedlot was and continues to be effectively isolated from down-gradient receptors. This occurs because of potentially strong sequestration by Fe- and Ca-phosphate mineral precipitation, with transport further mitigated by organic matter adsorption in wetlands surrounding the feedlot.

Phosphorus remaining at the feedlot. The total mass of phosphorus remaining at the feedlot can be estimated roughly from the dimensions of the groundwater and soil P anomaly. For groundwater, the average porosity of the saturated sediments above the underlying clay (0.15) was multiplied by the sediment volume (2.7 × 10⁵ m³) and the average concentration of P that exceeds the background value (0.1 mg/L). Results showed that only about 4 kg of P occur dissolved in sediment pores. A similar estimate can be carried out for the P sequestered within the soils. In this case, the volume of the sand and silt that lies above the clay till (3.6 × 10⁵ m³) is multiplied by both the dry bulk density (1,400 kg m³) and the average concentration of P above background (35 mg/kg), giving a result of 1.8 × 10⁴ kg of P. Therefore, only a very small fraction of the total P at the site occurs in groundwater (0.02%).

Unfortunately, no documentation exists on how many cattle were sheltered at the feedlot during its 30 years of existence, other than an average of about 500 head during the years of the operation, nor is there any record of the amount of manure removed from the site. Regardless, much of the manure would have been moved off-site and spread on nearby crop fields. Assuming that 500 cattle were present in the lot, on average, along with a yearly P deposition rate of 18 kg P per cow (Goolsby et al. 1999), then perhaps 2.7 × 10⁵ kg of P would have been generated, which is roughly 15 times more than presently sequestered at the site.

An additional consideration on the fate of phosphorus relates to the oxidation-reduction dynamics once all of the nitrate has been denitrified. Carbon continues to be transported to groundwater from decaying and leaching of vegetation at the surface, which will lead to increasingly reduced conditions in the shallow groundwater. The sequence of microbial-mediated oxidation-reduction processes proceeds from NO₃^– / N₂ to MnO₂(mineral) / Mn²⁺ to Fe(OH)₃ (mineral) / Fe²⁺ oxidation-reduction couples (e.g., Stumm and Morgan 1996). The oxidant with the highest thermodynamic potential is used initially, followed by the one with the next lower potential. Because only small concentrations of nitrite and manganese occur at the site, microbial reduction and dissolution of iron minerals which sequester P may occur once nitrate is completely used. As a result, this process may trigger a later mobilization of phosphorus.

Nutrient transport and downstream water quality. In addition to the focused study at the feedlot, the water quality within nearby Judicial Ditch 66 (Figure 2) was tracked during and prior to the geochemical characterization and monitoring at the abandoned feedlot. As observed elsewhere, a rise in phosphorus transport in runoff often occurs following restoration of wetlands (Aldous et al. 2005, 2007). In the case of Judicial Ditch 66, an increase of total P had been observed in ditch discharge flowing northward out of the Glacial Ridge National Wildlife Refuge following wetland restoration in the basin, although nitrate had decreased during the same period (Figure 9). Because of the strong sequestering of P at the feedlot, the observed transport of P from the ditch system was instead a likely result of the chemical reduction of soils and sediments in the recently restored wetlands. Newly saturated soils will lose much of their capacity for ferric adsorption, thereby releasing P to surface waters. This process is also consistent with the observed reduction in nitrate concentration (Figure 9), as nitrate is readily denitrified in reducing conditions (e.g., Korom 1992) brought on by soil saturation. Monitoring of Judicial Ditch 66 both up- and downstream from the feedlot suggested that aggregate mining and the disturbance from prairie restoration may account for the small downstream increase in P, rather than transport from the abandoned feedlot. Furthermore, restoration and a concomitant return to anaerobic conditions has likely helped mitigate excessive nitrate throughout the system.

Old, abandoned feedlots may serve as a source of nutrients that can degrade groundwater and downstream water quality. At the Crookston Cattle Company feedlot, which operated from 1970 through 1999 and abandoned in 2000, some nutrients remain and a former waste lagoon continues to serve as a source of nitrate. N isotopes and temporal monitoring, however, suggest that most nitrate has been denitrified. Phosphorus remains strongly fixed beneath the feedlot and the site is surrounded by soils that have even greater sequestration capacity. Diminishing organic matter and dissolved organic carbon occurred after cattle were removed from the feedlot, thereby mitigating continued transport of organic P. This indicates that P will likely be retained to a greater extent at the site, provided the land use/land cover do not change significantly on the site’s calcareous soils.

In the case of the Crookston Cattle Company’s landscape setting, results indicate that the partitioning of nutrients was similar between the soil B-horizon and groundwater, suggesting that site characterization may not have required costly installation of groundwater monitoring equipment. Instead, soil sampling and routine analysis would suffice to characterize the transport and storage of P. This suggestion, however, might only apply to sites with a similar shallow water table and deep B-horizon soils.

Results indicate that although N and P concentrations are not excessive in adjacent wetlands and do not appear to have been transported into downstream channels. Nonetheless, robust invasive vegetation is well established in the nutrient-rich soils surrounding the feedlot and would be difficult to reclaim to more natural conditions. Disturbance, even decades earlier, provided an avenue for establishment of non-native weeds, which was later and continues to be augmented by abundant nutrient resources. Ditch closure and return to a more temporally and spatially consistent, natural hydrological pattern following the surrounding site restoration did little to mitigate weeds, and therefore does not alone assure or provide a trajectory toward recovery.

The ability to assess thoroughly the outcome of the nutrients released to the surrounding environment is limited greatly by our lack of feedlot records, including the number of animal units held in the pens at various times, the nutrient concentration of excrement, and the approximate proportion of manure retained and removed from the site. Nonetheless, learning from the legacy nutrients at Crookston Cattle Company feedlot suggests that on-site investigation might be carried out to characterize the suitability of potential future feedlot sites. Locations could be selected that provide the best conditions for denitrification, which requires the presence of anoxic soils and labile organic carbon or other electron donors. Soils that are well drained and contain large concentrations of organic matter, calcium, and ferric iron, similar to those at the Crookston Cattle Company feedlot, would likely be those most favorable for mitigating nitrate transport and enhancing phosphorus sequestration.

All data used are presented in this published report or its accompanying supplemental files.

The authors wish to thank the U.S. Fish & Wildlife Service Plains & Prairie Potholes Landscape Conservation Cooperative (PPP-LCC, administered through the Red Lake Watershed District) and the North Dakota Water Resources Research Institute for financial support. We also appreciate The Nature Conservancy’s permission to access and monitor the former feedlot site. Several anonymous reviewers, along with Ariane Peralta, East Carolina University, especially, provided thorough reviews and suggestions for improving the manuscript.

Most funding for this project came from the U.S. Fish & Wildlife Service’s Plains & Prairie Potholes Landscape Conservation Cooperative agreement 301817J119M4, administered through the Red Lake Watershed District. Additional funding came from the North Dakota Water Resources Research Institute and the Harold Hamm School of Geology and Geological Engineering.

The authors have no competing interests to declare.

• Contributed to conception and design: PJG, PG

• Contributed to acquisition of data: PJG, PG

• Contributed to analysis and interpretation of data: PJG, PG

• Drafted and/or revised the article: PJG, PG

• Approved the submitted version for publication: PJG