Suppression of hypolimnetic methylmercury accumulation by liquid calcium nitrate amendment: redox dynamics and fate of nitrate

Abstract

Anoxia, specifically sulfate (SO₄²⁻) reduction, in the hypolimnia of stratified lakes is a prerequisite for monomethylmercury (HgCH₃⁺) formation in sediments and subsequent efflux into the water column. A whole-lake pilot project used liquid calcium nitrate (LCN) to suppress formation of HgCH₃⁺ by raising hypolimnetic redox in a dimictic, mesotrophic lake in Minnesota. A nearby lake of similar morphology and trophic status served as a reference. In the experimental lake, the HgCH₃⁺ concentration at the onset of LCN application was 0.34 ng/L and 0.30 ng/L at the point of nitrate (NO₃⁻) depletion. The NO₃⁻ concentration peaked at 17.7 mg/L (as ion) 16 days after application and declined to 0.5 mg/L at day 64. A principal NO₃⁻ sink was anaerobic oxidation of sulfidic compounds, but dissimilatory NO₃⁻ reduction to ammonium (NH₄⁺) may also have been a major sink. SO₄²⁻ at the time of the LCN dose was 0.99 mg/L and rose to 4.0 mg/L at the point of NO₃⁻ depletion (<0.5 mg/L). Thereafter, NO₃⁻-induced SO₄²⁻ enrichment of the hypolimnion induced HgCH₃⁺ to increase to 1.09 ng/L by SO₄²⁻ reduction. In both lakes, we observed strong redox functional relationships in the hypolimnia with unambiguous breakpoints (thresholds) for SO₄²⁻ reduction and sediment efflux of iron, manganese, orthophosphate, NH₄⁺, and HgCH₃⁺. Study results suggest that LCN application be done in frequent, small doses with redox monitoring to prevent SO₄²⁻ reduction until turnover. This method would minimize downstream NO₃⁻ loading at turnover while avoiding inadvertent SO₄²⁻ enrichment that can boost HgCH₃⁺ production in SO₄²⁻ limited lakes.

Key words:

• anoxia
• calcium nitrate
• denitrification
• hypolimnion
• methylmercury
• redox
• sulfide

Atmospheric deposition of mercury (Hg) is the root cause of widespread accumulation of methylmercury in freshwater fish tissue (Watras 2009). Sediment bacteria convert inorganic Hg under anoxic conditions into monomethylmercury (HgCH₃⁺), making it bioavailable to primary and secondary producers (Ullrich et al. 2001, Hamelin et al. 2011). Biomagnification makes HgCH₃⁺ highly toxic to top predators in food webs (Boening 2000). Elevated methylmercury content in fish tissue of 0.3 mg/kg wet weight causes designation of lakes as impaired or promulgation of fish consumption advisories (Watras et al. 1998, US EPA 2001, Monson and Heiskary 2008, Evers et al. 2011, Zananski et al. 2011).

Source control of Hg emissions is ultimately the most effective solution to atmospheric Hg pollution (Johansson et al. 2001, Morrison 2011), but Hg deposited to terrestrial areas will continue to load receiving lakes into the future, depending on site-specific factors (Scherbatskoy et al. 1998, Driscoll et al. 2007). Response of lakes to Hg loading is not instantaneous. Rather, there is potential for many years, perhaps decades, before Hg methylation reaches a steady state from mercury loading (Paterson et al. 2006). Effective Hg source control will depend on effective implementation of the Minamata Convention on Mercury, which itself will take decades to substantially reduce atmospheric Hg emissions (UNEP 2013). Consequently, Hg source control is unlikely to adequately remediate near- to intermediate-term Hg contamination problems for many Hg-impaired lakes that have received Hg from atmospheric loading.

The scale of Hg contamination severely limits a lake-by-lake approach to remediation of lakes impaired by Hg. A small subset of these lakes or reservoirs, however, will warrant the expense of site-specific remediation, particularly if technologies that suppress Hg methylation also suppress internal loading of nutrients in eutrophic lakes.

Anoxic hypolimnia can be important sources of HgCH₃⁺ in lakes (Eckley et al. 2005, Watras 2009). In these lakes, relief from hypolimnetic anoxia offers potential remediation of Hg-impairment (Beutel et al. 2008b, Todorova et al. 2009, Dent et al. 2014, Matthews et al. 2013).

Sulfate (SO₄²⁻) reducing bacteria (SRB) are the principal producers of HgCH₃⁺ in sediments (Compeau and Bartha 1985, Benoit et al. 2003, Effler and Matthews 2008, Corrales et al. 2011), but other bacteria, including iron (Fe)-reducing bacteria and methanogens, possess genes for Hg methylation (Gilmour et al. 2013, Parks et al. 2013). Elevated redox conditions at the sediment surface, too oxidizing for these reductive processes to occur, suppress HgCH₃⁺ formation (Regnell and Tunlid 1991, Boszke et al. 2003, DeLaune et al. 2004). Thus, concentrations of nitrate ion (NO₃⁻; Beutel et al. 2008b, Todorova et al. 2009, Shih et al. 2011) or dissolved oxygen (DO; Beutel et al. 2008b, Dent et al. 2014) above critical concentrations suppress HgCH₃⁺ formation by suppression of SRB activity in surficial sediments.

Engineering methods can elevate hypolimnetic redox by adding NO₃⁻ or DO. The design practice for DO is well-understood (Beutel and Horne 1999, Cooke et al. 2005, Gantzer et al. 2009b) and has also demonstrated suppression of HgCH₃⁺ production (Beutel et al. 2008b, Dent et al. 2014).

The so-called Riplox method injects NO₃⁻ in the hypolimnia to control internal loading of phosphate (PO₄³⁻; Ripl 1978, Willendbring et al. 1983, Noonan 1986). Although our engineering team was also engaged in hypolimnetic oxygen projects, we investigated NO₃⁻ addition because of its low infrastructure needs and potential to control both HgCH₃⁺ and internal nutrient loading. Almost all hypolimnetic oxygenation projects occur in large reservoirs for utility clients. There is a need for remediation technologies appropriate at small scales and for clients with limited capacity to operate remediation infrastructure.

Internal loading of PO₄³⁻ is largely a function of SO₄²⁻ and Fe reduction in surficial sediments (Mortimer 1942, Holmer and Storkholm 2001, Christophoridis and Fytianos 2006), which elevated NO₃⁻ will suppress. Todorova et al. (2009) observed suppression of HgCH₃⁺ by residual NO₃⁻ in the hypolimnion of Lake Onondaga; the source of NO₃⁻ was wastewater discharge. Production of HgCH₃⁺ returned after depletion of hypolimnion NO₃⁻. Subsequently, but after the present study, Matthews et al. (2013) demonstrated suppression of HgCH₃⁺ through injection of liquid calcium nitrate (LCN) to the hypolimnion of Lake Onondaga.

Peer reviewed literature or publically available engineering reports available at the time of the present study (2009–2010) did not provide a sufficient engineering design basis to assess the efficacy and practical application of NO₃⁻ addition for suppression of SRB activity in anoxic hypolimnia. We resolved, therefore, to conduct a lake-scale pilot project to investigate the design basis for HgCH₃⁺ management.

This study documents a pilot project that investigated NO₃⁻ addition to control internal loading of HgCH₃⁺ in the hypolimnion of a mesotrophic lake in Minnesota listed for Hg impairment. The study needed to address basic engineering issues:

1. evaluation of the efficacy of NO₃⁻ addition to suppress HgCH₃⁺ formation;

2. exploration of a rational basis for establishing NO₃⁻ dosing criteria;

3. development of dosing methods;

4. determination of dominant NO₃⁻ sinks and rate of NO₃⁻ demand;

5. evaluation of operational parameters indicative of treatment efficacy; and

6. exploration of regulatory concerns over NO₃⁻ toxicity and nutrient enrichment, which are intrinsic issues surrounding NO₃⁻ addition.

Study location

The 3 study lakes, Round, Ann, and Riley, are all located in the Riley Purgatory Bluff Creek Watershed District in the western suburbs of Minneapolis, Minnesota (Table 1). The water sources for all study lakes are suburban runoff and groundwater with no known industrial discharges. All study lakes are dimictic, stratifying immediately after ice thaws in late April or early May. All lakes have a May onset of hypolimnetic anoxia that continues until fall turnover. Lake basin morphology is similar, characterized by a deep (11–16 m) central basin bordered by a littoral fringe.

Table 1 Study lake descriptions. Addition lake descriptions available at http://www.dnr.state.mn.us/lakefind/index.html

Lake	County	Depth m	Area ha	Littoral %	Trophic Status	Latitude	Longitude
Round	Hennepin	11	13	74	Mesotrophic	44°59′04.02^″	93°31′29.68^″
Ann	Carver	14	45	41	Mesotrophic	44°52′16.55^″	93°33′25.74^″
Riley	Carver	16	120	38	Eutrophic	44°50′07.38^″	93°31′15.93^″

Table 2 Wet chemistry parameters and analytical methods applied in this study.

Parameter	Method
Methyl-mercury	EPA 1630
Orthophosphate	EPA 365.3
Ammonia	EPA 350.1
Nitrate	EPA 365.3
Manganese	EPA 6010B
Iron	EPA 6010B
Sulfate	EPA 300.0

Materials and methods

Round Lake was the focus of this study. It received a dose of LCN with the intent of suppressing HgCH₃⁺ production. Lake Ann served as a reference lake monitored with the same protocol as Round Lake. Riley Lake was lumped with Round Lake and Lake Ann in the correlation of sulfide (H₂S) odor with redox. We used 2008 (Jun–Oct) and 2009 (Apr–Oct) monitoring data for method development, as described below, and conducted the LCN application pilot study in 2010 (Mar–Oct). We used a Hydrolab DS5 multi-probe to obtain DO (ion selective electrode), redox (platinum band electrode), pH, conductivity, and temperature at 1.0 m intervals from surface to the bottom. We calibrated the redox probe (single point) to +428 mV in Zobell's solution at 25 C to the standard hydrogen reference electrode (SHE) scale. Field monitoring protocol allowed 3–5 minutes of equilibration of probes before recording values manually and to a data logger.

We used Van Dorn bottles to obtain wet samples from the surface, mid-metalimnion, and 1.0 m from the bottom (Table 2). Data reported as the hypolimnion are from samples 1.0 m from the bottom. We obtained all samples from a boat with the exception of the March 2010 HgCH₃⁺ sample obtained from a hole in the ice. Surficial sediments in both Round Lake and Lake Ann were flocculent. Starting in early summer, inadvertent impact of the bottom with the van Dorn bottle produced plumes of bubbles a few meters in diameter. Field crews were careful not to entrain sediments into near-bottom samples and to shift sampling location if there was contact with the bottom.

Production of H₂S within the hypolimnia of study lakes was monitored through depletion of SO₄²⁻ and by a field odor assessment of sample bottle contents using a scale of 1-nondetect (nose close to sample), 2-presence (detection with nose close to sample), 3-moderate (distinct sulfide odor near van Dorn bottle at opening), and 4-strong (distinct sulfide odor across the boat upon opening van Dorn bottle). This method is similar to the serendipitous sniff test (Barbash 2002) but developed independently for this study to distinguish more than presence or absence of sulfide in the sample. The field monitoring team always had a member with experience in the scale of comparison.

The NO₃⁻ application to Round Lake was a slug dose of LCN [Ca(NO₃)₂• 4H₂O], specific gravity 1.452. By mass, LCN total nitrogen (N) and total calcium (Ca) were 8.0 and 12.3%, respectively.

A weighted hose discharged LCN to the deepest point of the hypolimnion, pumped directly from tanker trucks. The small size of the pelagic area (3.64 ha) and simple, bowl-like morphology of Round Lake were ideal for this method of application. Two 20,412 kg (45,000 lb) tanker trucks of LCN each provided 3266 kg of NO₃⁻ (as ion) to the hypolimnion. The tanker trucks were cleaned before picking up the LCN for delivery to avoid contamination of unwanted compounds.

Regulatory concerns and NO₃⁻ depletion rates calculated from sediment oxygen demand (SOD) informed calculation of the LCN dose. The Minnesota Pollution Control Agency (MPCA) expressed concern that an NO₃⁻ ion concentration >30 mg/L could be toxic to invertebrates, and the US EPA had not promulgated a NO₃⁻ standard at that time. This project was informed instead by the promulgated Canadian standard of 13 mg/L NO₃⁻ (as ion; Environment Canada 2003), an order of magnitude lower than the lowest observed effect concentration (LOEC) toxicity threshold. The northern leopard frog (Rana pipiens) had a 56-day LOEC of 133 mg/L NO₃⁻, and a water flea (Ceriodaphnia dubia) had a 7-day LOEC of 189 mg/L NO₃⁻. Acute toxicity thresholds (96 hr-LC₅₀, 120 hr-LC₅₀) are at least an order of magnitude higher, except for the caddis fly (Hydropsyche occidentalis), which has a 120-hour LC₅₀ of 290 mg/L NO₃⁻.

The final criteria developed for an LCN dose maximum were 2-fold: (1) the initial NO₃⁻ concentration needed to be less than the lowest known LOEC, and (2) the NO₃⁻ concentration would not exceed the MPCA informal threshold of 30 mg/L NO₃⁻ for more than 30 days. Per Canadian standards, these criteria would fall well below known LOEC for NO₃⁻.

Determination of the initial dose provisionally assumed that the hypolimnion of Round Lake is well mixed. This assumption simplified dose calculation by avoiding uncalibrated modeling of plume dispersion and dissolution rates from the pool of LCN resting on the lake bottom. The nominal initial dose concentration would be the mass of the NO₃⁻ dose divided by the hypolimnion volume (253,354 m³) determined from lake bathymetry for depths <3.8 m, which is the approximate location of the top of the thermocline.

NO₃⁻ is an alternative electron acceptor to O₂ for bacterial respiration. There is a well-known equivalence of NO₃⁻ demand to O₂ demand in wastewater engineering (Metcalf and Eddy 2003); therefore, hypolimnetic oxygen demand (HOD) can be used to approximate NO₃⁻ demand. Because 77.4% of NO₃⁻ is oxygen on a mass basis, we calculated the hypolimnetic NO₃⁻ demand as HOD/0.774.

Determination of HOD was problematic. Although DO profile data were available from late April or early May through September in 1996, 1997, 2001–2004, 2008, and 2009, in all years, the hypolimnion was anoxic (<1.0 mg/L DO) from the onset of spring monitoring. Calculation of HOD from DO depletion curves was not possible.

Direct measurement of HOD as the sum of SOD and suspended community oxygen demand (SCOD) provides an alternative to oxygen depletion curves. In 2009, the SOD was 2.16 g/m²/d after fall turnover (temperature 15 C), using the EPA in situ method of (diver-placed) isolation chambers on the sediment surface (Murphy and Hicks 1986, US EPA 2009). Site conditions prevented deployment of the diver support boat, forcing deployment of isolation chambers from the shore in 4.0 m of water rather than 11.0 m at deepest depth. This depth was in the thermocline, but anoxia extends to the top of the thermocline in Round Lake. The 50^th percentile of July–August DO (n = 26) at 4.0 depth in 1997, 2001, 2002, 2003, 2008, and 2009 was 0.5 mg/L, with only the 75% percentile >2.0 mg/L. At depths of 10 m or greater (n = 27) during the same period, all DO values were <0.5 mg/L. Given the stronger anoxia at deepest depth, the SOD may be higher than at 4.0 m. The SCOD was assumed to be 1.04 g/m²/d O² based on a 2010 winter diel curve analysis (Odum and Hoskins 1958) study completed on nearby lakes of similar trophic status in the Bluff Creek Watershed District (CH2M HILL unpubl. data). The hypolimnion area (A_H) under the 3.8 m depth curve is 47,299 m². Thus HOD = A_H(SCOD + SOD)/d = 151 kg/d O₂. Expressed as NO₃⁻ demand, HOD_asNO3 = 195 kg/d NO₃⁻. Expressed as area, this rate is 4.1 g/m²/d NO₃⁻.

Two truckloads of LCN provided a total dose 6532 kg NO₃⁻. Divided by the hypolimnion volume, the resulting nominal NO₃⁻ concentration in the hypolimnion would be 26 mg/L. Per the areal depletion rate, NO₃⁻at this concentration would fully deplete in ∼43 days. A third truckload would have raised the NO₃⁻ concentration to >30 mg/L. Assuming a linear depletion rate, 2 truckloads of LCN would meet the Canadian criterion within 3 weeks. Finally, a longer period would run the risk of not achieving complete NO₃⁻ depletion prior to thermal erosion of the hypolimnion in September. Complete depletion of NO₃⁻ was essential to our study goals, and therefore the dose was 6532 kg NO₃⁻ (2 truckloads of LCN).

Graphic analysis of data used Excel and Kaleidagraph software. Nonparametric statistical analyses of data employed Wilcoxon–Mann–Whitney rank sum tests using Kaleidagraph software.

Table 3 Free energy for selected potentially competing equations for nitrate reduction at standard conditions.

Competing Reactions for Nitrate	ΔfG°, kJ/mol
Heterotrophic (acid-consuming)
CH3COO^− + 0.8 NO3^− + 1.8 H^+ → CO2 + 0.4 N2 + 2.4 H2O	−466
CH4 + 1.6 NO3^− + 1.6 H^+ → CO2 + 0.8 N2 + 2.8 H2O	−188
Chemolithotrophic (acid-producing)
H2S + 1.6 NO3^− → 0.8 N2 + 0.8 H2O + SO4^2− + 0.4 H^+	−713
FeS2 + 3 NO3^− + 2 H2O → Fe(OH)3 + 2 SO4^2− + 1.5 N2 + 1H^+	−1185
Fe^2+ + 0.2 NO3^− + 2.4 H2O → Fe(OH)3 + 0.1 N2 + 1.8 H^+	−11
Dissimilatory Nitrate Reduction – DNRA (acid neutral)
CH2O + 0.5 NO3^− + 0.5 H2O = 0.5 NH4^+ + HCO3^−	−316
H2S + NO3^− + H2O = NH4^+ + SO4^−2	−393

Gibbs free energy (kJ/mol) for likely completing reactions for NO₃⁻ (Table 3) were obtained from thermodynamic references at standard conditions (Woods and Garrels 1987, Lide 2007) using: (1)

Results

On 15 June 2010, 2 semi-trailer truckloads discharged the LCN dose to the hypolimnion of Round Lake. The NO₃⁻ hypolimnion concentration peaked 15 days after application at 17.7 mg/L NO₃⁻ (Fig. 1). By 18 August (64 d), the NO₃⁻ concentration had dropped to 0.34 mg/L. Depletion of NO₃⁻ from the concentration peak was linear at a rate of 102 kg/d, which is 2.8 g/m²/d, 67% of the estimated depletion rate.

Figure 1 Methylmercury and nitrate ion concentrations in the Round Lake hypolimnion, 2010. Vertical line is from date of and just prior to LCN addition.

The HgCH₃⁺ hypolimnetic concentration of Round Lake was 0.35 ng/L before the application of LCN and remained at this value with minor fluctuations until depletion of NO₃⁻ (Fig. 1). Addition of LCN produced SO₄²⁻ (Fig. 1). After depletion of NO₃⁻, the HgCH₃⁺ concentration rose to 1.09 ng/L by 15 September (Fig. 1). DO in the hypolimion during the study was <0.1 mg/L, leaving no other oxidation mechanism for sulfidic compounds (e.g., HS⁻, H₂S, FeS, FeS₂) other than anaerobic oxidation via NO₃⁻.

Lake Ann hypolimnetic DO was <0.1 mg/L May–September 2010. Redox values tracked production of HgCH₃⁺ and geochemical shifts of solute species. HgCH₃⁺ increased monotonically from 0.085 to 0.55 ng/L through the same period in inverse proportion to redox values (Fig. 2). Decreasing redox values tracked SO₄²⁻ reduction and accompanying solubilization of Fe, Mn, and Ortho-P from the sediments (Fig. 3). A similar redox dynamic occurred in the hypolimnion of Round Lake (Fig. 4).

Figure 2 Lake Ann HgCH₃⁺ and redox, 2010.

There were significant differences between redox values and H₂S odor intensity (Fig. 5) that corresponded to redox solubilization thresholds observed in Round Lake and Lake Ann (Fig. 3). The median redox value of +53 mV for samples with a strong H₂S odor was significantly different (P = 0.015) from samples with a moderate odor (+70 mV). The median redox value of +263 mV for samples of weak or no odor was significantly different than moderate odor samples (P < 0.0001) and strong odor samples (P < 0.0001). By contrast, there was no statistically significant difference between median DO of strong and moderate sulfide odors (P = 0.48) or moderate and weak odors (P = 0.9; Fig. 5). In the trivial case of no odor, there was a significant difference (P < 0.012) between median redox values of other odor categories. Across all lakes, +60 mV was the approximate threshold below which strong H₂S odors where detected and median HgCH₃⁺ concentrations were significantly higher (Fig. 6). Sharp increases in Ortho-P solubilization began in the upper quartile (∼+200 mV) of the moderate H₂S odor redox range.

Figure 3 Relationship between redox-active chemical species and redox values in Lake Ann, 2010.

Figure 4 Relationship between redox-active chemical species and redox values in Round Lake, 2010.

Figure 5 Statistical distribution of sulfide odor detection for redox values (left) and dissolved oxygen (right). Sample sizes: strong, 41; moderate, 16; weak, 7; none, 27.

Both Lake Ann and Round Lake hypolimnia were strongly anoxic. Lake Ann redox was in the manganese (Mn) and Fe reduction redox range from March to early July when it dropped into SO₄²⁻-reducing conditions, where it remained until early September (Fig. 7). Round Lake redox conditions were already in the SO₄²⁻-reducing range when LCN was placed in the hypolimnion. Redox values rose from +45 mV prior to application on 15 June, to +85 mV on 16 June, to +138 mV by 17 June. By 01 July, redox values were in a range (>+300 mV) typically associated with oxic conditions (Fig. 7). By mid-August, redox values dropped to SO₄²⁻-reducing values with complete depletion of NO₃⁻.

The dynamics of SO₄²⁻ production in Lake Ann and Round Lake indicated that sulfidic compounds were important, possibly dominant, electron donors in NO₃⁻ reduction (Fig. 7). Whereas there was steady depletion of SO₄²⁻ with dropping redox in Lake Ann, in Round Lake SO₄²⁻ rose from 0.99 mg/L at the time of LCN addition to 4.0 mg/L just after depletion of NO₃⁻.

Assuming that SO₄²⁻ concentrations were uniform in the hypolimnion, SO₄²⁻ mass increased by 760 kg. Per H₂S oxidation by NO₃⁻ (Table 3), as much as 784 kg NO₃⁻ was consumed, or 12% of the NO₃⁻ dose. This calculation assumes a uniform hypolimnetic NO₃⁻ concentration of 0.34 mg/L (86 kg NO₃⁻ residual mass).

After addition of NO₃⁻, soluble, but not total, Fe temporarily decreased in Round Lake (Fig. 8), but there was no significant effect on Mn concentration. In Lake Ann, soluble and total Fe and Mn were equal and increased throughout the study (Fig. 8). Assuming that the drop in Round Lake hypolimnetic Fe(II) was due to uniform oxidation by NO₃⁻ throughout the hypolimnion, ∼963 kg of Fe(II) oxidized to Fe(III). Oxidation of Fe(II) in the hypolimnion would account for up to 214 kg NO₃⁻ (Table 3), ∼3% of the NO₃⁻ dose.

Figure 6 Hypolimnion HgCH₃⁺ concentrations in relation to redox value in all lakes, 2009–2010.

Figure 7 Hypolimnion sulfate and redox values in Lake Ann and Round Lake, 2010. Vertical line is date of LCN addition to Round Lake.

Figure 8 Hypolimnion Fe and Mn in Lake Ann (left) and Round Lake (right). Vertical line is day of LCN addition in Round Lake.

There may have been dissimilatory NO₃⁻ reduction to NH₄⁺ (DNRA) in Round Lake, which is thermodynamically favorable in both heterotrophic and autotrophic bacteria (Table 3). The NH₄⁺ concentration increased 2.0 mg/L on 16 June to 5.5 mg/L on 18 August. Assuming a uniform distribution of NH₄⁺ concentration in the hypolimnion, on a mass ratio basis of 3.4 g NO₃⁻:1.0 g NH₄⁺, the 881 kg increase in hypolimnetic NH₄⁺ would consume 3034 kg of NO₃⁻, which is 46% of the dose.

For Round Lake hypolimnetic NH₄⁺ concentrations after 15 June, there was a significant difference (P = 0.019) between the 2010 median N of 4.0 mg/L (min 2.4, max 6.2, n = 8) and the 2008–2009 median N of 2.2 mg/L (min 0.4, max 4.6, n = 9). By comparison, there was no significant difference (P = 0.08) between the median concentration in Lake Ann (1.6 mg/L) and Round Lake (2.4 mg/L). NH₄⁺ concentrations in both lakes had similar responses to redox (Fig. 9) that followed redox-active solubilization of Ortho-P, Fe, and Mn (Fig. 3 and 4). The hypolimnetic NH₄⁺ concentration rose in Round Lake from 4.4 mg/L on 18 August to 6.3 mg/L on 16 September, but there was no NO₃⁻ to produce NH₄⁺ during this period.

Discussion

After addition of LCN to Round Lake, formation of HgCH₃⁺ in the hypolimnion did not increase until depletion of NO₃⁻. By contrast, HgCH₃⁺ production continued unabated throughout the same period in Lake Ann. There are no baseline HgCH₃⁺ data for either lake to evaluate the potential impacts of interannual variability. Suppression of HgCH₃⁺ formation by NO₃⁻ addition was consistent with findings by Todorova et al. (2009) and Matthews et al. (2013) in Lake Onondaga, New York. NO₃⁻ concentrations >0.5 mg/L suppressed HgCH₃⁺ production in a cold (<10 C) hypolimnia in both the current study and that of Lake Onondaga. In Lake Onondaga, mechanisms of NO₃⁻ control of HgCH₃⁺ formation were suppression of SRB activity, decrease in methylation/demethylation ratios, and increased sorption to Fe and Mn hydroxides.

Figure 9 Lake Ann and Round Lake NH₄⁺ concentrations as function of redox. High-redox, high-NH₄⁺ values for Round Lake all occurred after NO₃⁻ dosing and prior to NO₃⁻ depletion.

Data from this study implicate SO₄²⁻ reduction in the production of HgCH₃⁺. Strong H₂S odors were unambiguously associated with the lowest recorded redox values (Fig. 3, 4, and 5) and with HgCH₃⁺ production (Fig. 2 and 6).

Addition of NO₃⁻ immediately raised redox values above those associated with SO₄²⁻ reduction (Fig. 3, 4, and 7). The mechanism of redox elevation was NO₃⁻ oxidation of sulfidic compounds, as seen in the increase in SO₄²⁻ concentrations after the introduction of NO₃⁻ in Round Lake. Lake Ann, by contrast, exhibited monotonic SO₄²⁻ depletion and decrease of redox values during the same period.

Nitrate sinks

Results from the present study and thermodynamics support anaerobic sulfide oxidation as an important, possibly primary, sink for NO₃⁻ depletion. The NO₃⁻ dose of 6532 kg was more than sufficient to produce the apparent Round Lake hypolimnetic increase of 760 kg SO₄²⁻. Free energy for potential denitrification stoichiometry in the hypolimnion reveals that chemolithotrophic bacterial (utilizing inorganic C for respiration) oxidation of H₂S and FeS₂ is energetically favored over heterotrophic bacterial (utilizing organic C for respiration) reactions (Table 3). Organic carbon, therefore, may not be a primary sink for NO₃⁻.

The pH data support anaerobic H₂S oxidation via NO₃⁻. Chemolithotrophic denitrification produces acid (Table 3), DNRA neither produces nor consumes acid, and heterotrophic denitrification consumes acid. There was a drop in pH of 0.2–0.3 throughout the hypolimnion from LCN addition (15 Jun) until depletion of NO₃⁻ (18 Aug). Lake Ann did not undergo a midsummer depression in pH. There was a significant difference (P < 0.0001) between the Round 2010 median hypolimnetic pH (6.4) and the historic median (7.8) during 16 June–18 August 1997, 2001, 2002, 2003, 2004, 2008, and 2009.

DNRA may also have been an NO₃⁻ sink. Under conditions of NO₃⁻ limitation, bacteria cell growth yield and free energy available for respiration is higher for DNRA than for heterotrophic denitrification (Strohm et al. 2007). Four Round Lake hypolimnion samples of NH₄⁺ were 2.6 mg/L or greater at redox values of +175 mV or higher (Fig. 9) on dates between NO₃⁻ dose and depletion. Otherwise, all high concentration NH₄⁺ values followed the asymptotic functional relationship with low redox values exhibited by Fe, Mn, and Ortho-P (Fig. 3, 4, and 9).

Although results suggest DNRA, it is difficult to distinguish in Round Lake from similar hypolimnetic NH₄⁺ efflux in Lake Ann, in which there was no NO₃⁻ to drive DNRA. In Lake Ann, patterns of change in hypolimnion soluble and total Fe and Mn were identical throughout the study period (Fig. 8), indicating that the source for both was solubilization from sediments. In Round Lake, the drop in soluble Fe as total Fe increased might suggest NO₃⁻ oxidation of Fe(II) (Benz et al. 1998). Round Lake hypolimnetic Mn(II) did drop slightly concurrently with the drop in Fe(II), but oxidation of Mn(II) is strictly aerobic (Frankel and Bazylinski 2003, Burgin and Hamilton 2008) and can be ruled out. Co-precipitation of Mn(II) with Fe(III)is the likely cause of Mn(II) loss (Davison 1993, Hongve 1997). No storm or other high wind velocity event occurred that could have entrained hypolimnion waters and oxidized Fe or Mn. Regardless of mechanisms, results and thermodynamics (Table 3) indicate anaerobic Fe(II) oxidation was a minor sink for NO₃⁻.

Redox dynamics

Production of HgCH₃⁺ and in Lake Ann occurs as SO₄²⁻ nears depletion at low redox values (Fig. 2 and 3). There is an obvious inverse relationship between SO₄²⁻ depletion and solubilization of Fe, Mn, and Ortho-P (Fig. 3). Using Fe as a model, reduction of SO₄²⁻ produces H₂S, lowering redox values (Fig. 5). In turn, H₂S will abiotically reduce Fe and Mn, reducing competition to SRB from Fe- and Mn-reducing bacteria, which produces more H₂S, reducing yet more Fe and Mn, and so on, until depletion of SO₄²⁻ limits H₂S production. Observed Fe, Mn, and Ortho-P solubilization is a classic positive feedback relationship. In Lake Ann, hypolimnetic SO₄²⁻ reached maximum depletion by early August (Fig. 7), creating enriched H₂S conditions at the sediment surface that drove HgCH₃⁺ production for the rest of the summer (Fig. 2). Based on these data, observed HgCH₃⁺ production was apparently part of a larger redox system characterized by strong depletion of SO₄²⁻ and functionally linked to nutrient and metals mobilization from sediments. In Round Lake, the sharp jump in HgCH₃⁺ concentrations (0.35–1.1 ng/L) occurred after depletion of NO₃⁻ and as SO₄²⁻ reduction recommenced (Fig. 1 and 7). Thus, the NO₃⁻ dose to the hypolimnion apparently had the ultimate effect of stimulating HgCH₃⁺ production via SO₄²⁻ enrichment of the hypolimnion by NO₃⁻ oxidation of sulfidic compounds. Experimental design required NO₃⁻ depletion prior to turnover to observe NO₃⁻ sinks. To suppress hypolimnetic HgCH₃⁺ production, however, study findings indicate that it is essential either to maintain hypolimnetic NO₃⁻ concentrations above a minimum value until turnover elevates hypolimnetic DO or to prevent hypolimnetic SO₄²⁻ reduction by other means (e.g., oxygenation or chemical oxidant injection).

Evidence from this study and the literature indicate sediment SRB rapidly utilize NO₃⁻ as a terminal electron acceptor. Application of NO₃⁻ to sewers is a well-known method to suppress odors via NO₃⁻ oxidation of H₂S (Mathioudakis et al. 2006), where it must act rapidly to be effective. In a laboratory simulation of sewer mains, Mohanakrishnan et al. (2009) observed induction of anaerobic H₂S oxidation within 10 hours of exposure to NO₃⁻. In Round Lake, redox increased at an approximate linear rate of +1.9 mV/h in the first 48 hours after application, which demonstrates a similarly rapid system response to NO₃⁻.

Nitrate dose

Timing of the NO₃⁻dose was more than a month later than intended. Ice-out in early May presented the first opportunity to dose NO₃⁻, but logistical problems delayed application. It is not clear if timing would have changed results because limited winter data suggest that Round Lake may be permanently anoxic at deepest depths.

The observed rate of NO₃⁻ depletion of 2.8 g/m²/d was 67% of the estimated rate, suggesting that the dose was conservative. Potential causes of the underestimated error merit attention. SCOD may have been lower than assumed, isolation chamber placements in the SOD study may not have accurately captured overall SOD, or the difference may be an artifact of the isolation chamber method itself. Mixers circulate water in the chambers at a mean velocity of 3 cm/s. Velocity of water over sediments increases observed SOD by shear of diffusion boundary layers (Murphy and Hicks 1986, Gantzer et al. 2009b). There was no measurement of benthic boundary layer velocities in Round Lake. If seiche-driven velocities across the bottom averaged <3 cm/s, then the actual SOD may be lower than measured SOD.

The present study partially achieved a central goal of accurate determination of the NO₃⁻ dose. Matthews et al. (2013) determined dosing criterion by analysis of NO₃⁻ depletion curves observed in the hypolimnion of Lake Onondaga. Source of the NO₃⁻ was discharge of nitrified effluent from a regional wastewater treatment facility. Where an NO₃⁻ depletion rate is unavailable, the SOD method of determining the NO₃⁻ dose seems to be conservative.

From a management perspective, it is essential to know how much NO₃⁻ is sufficient to suppress formation of HgCH₃⁺. Although NO₃⁻ concentrations were >0.5 mg/L (15 Jun–18 Aug), redox values were substantially greater than values associated with HgCH₃⁺ production and detection of H₂S odors (Fig. 1, 5, 6, and 7). Maintenance of NO₃⁻ at 0.5 mg/L or higher, however, is not necessarily high enough to prevent formation of HgCH₃⁺.

Rates of SO₄²⁻ reduction are highly temperature dependent, approximately doubling from 0 to 12 C and every 10 C thereafter (Ingvorsen et al. 1981). During the study, the temperature in the hypolimnion of Round Lake was approximately 8 C. At higher water temperatures, higher NO₃⁻ concentrations in the water column may be necessary to suppress HgCH₃⁺. In addition, SO₄²⁻ in Round Lake and Lake Ann was <5 mg/L. Lakes with higher SO₄²⁻ concentrations may require higher NO₃⁻ concentrations to suppress SRB.

These qualifications suggest that the NO₃⁻ concentration needed to suppress SRB, and thus HgCH₃⁺ production, is likely be specific to geochemistry and physics of individual lakes (e.g., weak vs. strong seiches, sediment pore water vertical transport via ebullition, stratification dynamics, SOD, etc.). Mass flux of NO₃⁻ (or O₂) from the water column into surficial sediments, whether diffusive or advective in nature, must overcome mass flux of H₂S from sediments to the water column to maintain an oxidizing environment in the lake bottom water. A “rule of thumb” NO₃⁻ concentration to suppress SRB is not advisable.

Redox monitoring for methylmercury management

Redox monitoring may aid determination of NO₃⁻ concentration thresholds critical for lake managers to suppress HgCH₃⁺ production. Redox probes figured prominently in assessing geochemical state conditions within the hypolimnia of Round Lake and Lake Ann and merit detailed discussion. A redox probe is a millivolt meter. Snoeyink and Jenkins (1980) contend that a millivolt meter measures little, if anything, of significance in water chemistry. The US Geological Survey states that use of redox electrodes in water quality monitoring “is not recommended in general because of the difficulties inherent in its theoretical concept and its practical measurement” (Nordstrom and Wilde 2005). A redox probe does not measure water chemistry; rather it measures the effect of water chemistry on the millivolt probe, namely electron flow between redox couples; therefore, redox values have little meaning without corroborating data.

Interpretation of redox monitoring is further complicated by a wide range of response times across various probe types (Jang et al. 2005). Moreover, dependent on time of exposure, the drift in probe readings is sufficient to create large differences between in situ and manually deployed probes of the same type and in the same location (YSI Environmental 2005). The technical literature further confounds interpretation of redox data when it fails to state reference electrodes and calibration points.

Despite these issues, redox probes are important tools in wastewater operations as inputs to automated reactor controls for denitrifying systems (Wareham et al. 1993b, Demoulin et al. 1997, Zhao et al. 1999). Operational research establishes reliable empirical relationships between redox operating points and process goals. The NO₃⁻ breakpoint is a redox value associated with the disappearance of NO₃⁻ commonly and reliably used for real-time controls in wastewater operations (Wareham et al. 1993a, Chen et al. 2002, Kim et al. 2004).

These principles extend to hypolimnion management. Mere anoxia does not necessarily indicate HgCH₃⁺ production or SO₄²⁻ reduction (Fig. 3 and 4), both of which functionally relate to redox and not DO (Fig. 2 and 5). Redox is the only real-time monitoring tool that can assess state conditions within the hypolimnion that generate or suppress HgCH₃⁺.

As demonstrated by this study, there are clear breakpoints for HgCH₃⁺ production and SO₄²⁻ reduction. A positive feedback process as function of redox characterizes solubilization of Mn, Fe, and Ortho-P production (Fig. 3 and 4). The probable feedback mechanism comes from abiotic reduction of Mn(IV) and Fe(III) by H₂S produced in deeper sediment layers (Thamdrup et al. 1993), which then removes substrate competition for SRB, which then produce more H₂S/HS^−, and so on, until SO₄²⁻ is close to depletion (Fig. 3 and 4). Redox monitoring would allow real-time management of NO₃⁻ dosing to prevent HgCH₃⁺ production by keeping the hypolimnion above the HgCH₃⁺ production breakpoint or higher (e.g., redox values such as the SO₄²⁻ reduction breakpoint), preventing positive feedback.

Per the redox monitoring protocols of this study, the HgCH₃⁺ production breakpoint was apparently +60 mV (Fig. 6), suggesting that the HgCH₃⁺ production redox breakpoint may be higher. Detection of mild H₂S⁻ odor began at approximately +200 mV and strong odor at about +130 mV. These numerical redox values will change with differing redox protocols of a particular study and the mix of redox-active solutes in a given hypolimnion. What is unlikely to change is the existence of functionally similar breakpoints for HgCH₃⁺ production. Lake or reservoir managers can establish protocol and site-specific redox values for these breakpoints to monitor or manage water quality.

Biogeochemical system dynamics

The asymptotic increase in solubilization of NH₄⁺, Ortho-P, Fe(II), and Mn(II) as function of drop in redox merits closer attention (Fig. 3, 4, and 9). As a function of time, the increase in these solutes is exponential (Fig. 10). As Lake Ann hypolimnetic anoxia intensified over the summer, there was a nonlinear increase of redox-active solutes.

Figure 10 Lake Ann hypolimnetic Fe(II), NH₄⁺, and Ortho-P concentrations as function of time. Curve fits for time t (days): NH₄⁺ = 1 + 0.085e^0.35t (R² = 0.99); Fe(II) = 150 + 33e^0.27t (R² = 0.97); Ortho-P = 110 + 10e^0.36t (R² = 0.98).

The increases in Fe(II) and Ortho-P concentrations follow established trends in anoxic hypolimnia as a consequence of sediment–water interaction (Mortimer 1942). Insoluble Fe(III) PO₄³⁻ dissolves and releases PO₄³⁻ previously bound in Fe(OH)₃ complexes. The positive feedback dynamics of this process as controlled by SO₄²⁻, such as in the present study, have been observed since the mid-20^th century (Hupfer and Lewandowski 2008).

An explanation for the same pattern of NH₄⁺ efflux is not as straightforward because there is no shift in redox or solubility of N species. NH₄⁺ is highly soluble, and there was obviously no pool of NO₃⁻ in sediments that might reduce to NH₄⁺ via DNRA. A strong linear correlation (R² = 0.97) between hypolimnetic Fe(II) and NH₄⁺ seems to link Fe(III) reduction and Fe(II) efflux to NH₄⁺ effluent (Fig. 10).

A reasonable hypothesis for the NH₄⁺ efflux from sediments is ion exchange. A large fraction of NH₄⁺ in sediments is adsorbed to organic material and highly exchangeable (Rosenfeld 1979, Seitzinger et al. 1991). Fe(II) and Mn(II) are divalent cations produced by low redox conditions that will displace loosely bound monovalent cations such as NH₄⁺. Thus, the strong correlation of Fe(II) and Ortho-P efflux to NH₄⁺ efflux (Fig. 10) is probably causal rather than coincidental. Investigation of this hypothesis would entail measurement of sediment pore-water chemistry under study conditions.

The potential insight that NH₄⁺ efflux provides to HgCH₃⁺ efflux lies in similar redox-influenced ion exchange dynamics. Mass effects and charge are fundamental to ion exchange. When salinity is low, small increases in salinity dramatically increase efflux of NH₄⁺ from sediments (Hou et al. 2003). Like ammonium, HgCH₃⁺ preferentially binds to organic matter in sediments (Feyte et al. 2010). As monovalent cations, both are subject to similar ion-exchange dynamics. Salinity, as measured by specific conductivity (mS/cm), increased in the hypolimnion of Lake Ann as redox decreased. From +353 to +108 mV, salinity significantly (P < 0.002) increased by 1.9%, and by 2.8% (P < 0.0002) from +108 to +38 mV. The increase in salinity may seem small. Assuming a standard conversion factor of 0.67 (mg/L)/(mS/cm), these changes in salinity are 6.2 and 9.3 mg/L, respectively. Note the salinity concentration scale is parts per million, and the scale for HgCH₃⁺ is parts per trillion; a concentration difference on the order of 10⁶ has a powerful effect on ion exchange. If ion exchange substantially drives NH₄⁺ efflux at low redox, then similar physical chemistry is likely significant to HgCH₃⁺ efflux.

There seems to be an operational balance point for LCN application. Although there are legitimate concerns from the regulatory community that excess hypolimnion NO₃⁻ at fall turnover would flow downstream, evidence from this study indicates the need to maintain a sufficient hypolimnetic NO₃⁻ residual until fall turnover raises DO to the sediment surface. Without this residual, the net effect of LCN treatment in waters where HgCH₃⁺ production is SO₄²⁻ limited might be to enrich, rather than suppress, hypolimnetic HgCH₃⁺ concentrations. NO₃⁻ oxidation of sulfidic minerals increases the hypolimnetic SO₄²⁻ concentration, stimulating SRB activity if not suppressed by NO₃⁻ or DO. Hypolimnetic redox monitoring and small, frequent doses of LCN seem to be a reasonable management method to maintain this balance.

Although only peripherally within the scope of the present study, LCN control of Ortho-P efflux from sediments merits brief mention. Redox dynamics drive metal (Fe, Mn), nutrient (Ortho-P, NH₄⁺), and HgCH₃⁺ efflux from sediments. Maintaining the hypolimnion above a critical redox breakpoint will suppress these effluxes, strongly evidenced by hypolimnetic oxygenation projects (Beutel et al. 2008a, 2008b, 2014, Gantzer et al. 2009a, Dent et al. 2014). From a management perspective, redox controls sufficient to suppress HgCH₃⁺ would benefit nutrient and metals management.

This study demonstrated that dosing of NO₃⁻ suppresses HgCH₃⁺ production within an anoxic hypolimnion. Maintaining NO₃⁻ below known LOEC values while suppressing HgCH₃⁺ production was readily accomplished. Whether suppression of hypolimnetic HgCH₃⁺ formation translates into reduced Hg content of fish tissue is an important question that future studies should address. For lakes where HgCH₃⁺ of hypolimnion origin is an important source of Hg in fish tissue, NO₃⁻ dosing of the hypolimnetic merits investigation as a means to remediate Hg impairment.

Funding

This study was funded by the Riley Purgatory Bluff Creek Watershed District.