Mercury speciation in a coal-fired power plant plume: An aircraft-based study of emissions from the 3640 MW Nanticoke Generating Station, Ontario, Canada


  • Daniel A. Deeds,

    Corresponding author
    1. Department of Atmospheric and Oceanic Sciences, McGill University, Montreal, Quebec, Canada
    • Corresponding author: D. A. Deeds, McGill University, 945 Burnside Hall, 805 Sherbrooke St. West, Montreal, QC H3A 0B9, Canada. (

    Search for more papers by this author
  • Catharine M. Banic,

    1. Air Quality Research Division, Environment Canada, Toronto, Ontario, Canada
    Search for more papers by this author
  • Julia Lu,

    1. Air Quality Research Division, Environment Canada, Toronto, Ontario, Canada
    2. Department of Chemistry and Biology, Ryerson University, Toronto, Ontario, Canada
    Current affiliation:
    1. Department of Chemistry and Biology, Ryerson University, Toronto, Ontario, Canada
    Search for more papers by this author
  • Sreerama Daggupaty

    1. Air Quality Research Division, Environment Canada, Toronto, Ontario, Canada
    2. Now retired
    Search for more papers by this author

  • 4

    Now retired.


[1] Coal-fired power plants are one of the principal sources of mercury to the atmosphere. The form this mercury takes is the predominant factor determining its fate after emission. Recent ground-level field and modeling studies suggest that oxidized mercury in stack emissions is converted into elemental mercury in the plume. We present here aircraft-based plume mercury measurements taken by Environment Canada in 2000 at the Nanticoke Generating Station as part of the Health Canada Toxic Substances Research Initiative Metals in the Environment Research Network. Although the average mercury speciation observed in the Nanticoke plume (82% Hg0, 13% Hg(II)(g), 5% Hg(P), by mass) appears to be distinct from the average mercury speciation in the Nanticoke stacks (53% Hg0, 43% Hg(II)(g), 4% Hg(P)), we find that the in-plume elemental mercury concentrations as a whole can be explained by plume dilution after emission. The discrepancy between in-stack and in-plume Hg(II) concentrations is statistically significant, yet is not associated with a transformation of Hg(II) to Hg0. Sampling biases associated with the differing techniques used to measure Hg(II) in-stack and in-plume may reconcile the concentration discrepancy without invoking novel chemical reactions or physical processes. Although the mercury speciation of the Nanticoke plume influences local mercury deposition, the majority of the mercury emitted is transported out of the surrounding area.

1 Introduction

[2] Coal combustion is one of the principal sources of anthropogenic mercury to the environment, accounting for around 30–45% of new anthropogenic mercury emissions to the atmosphere annually [United Nations Environment Programme, 2008; Pirrone et al., 2010]. Mercury emitted from coal-fired power plants is predominately gaseous elemental mercury (GEM) and gaseous oxidized mercury (GOM, Hg(II)(g)) with a small contribution from particle-bound mercury (Hg(p)) [Pacyna et al., 2006]. Elemental mercury is relatively insoluble and inert, with an estimated atmospheric lifetime of 0.5–1.5 years [Bergan et al., 1999; Seigneur et al., 2006] and as such is subject to long-range atmospheric transport on a continental to global scale. Atmospheric GEM is converted through chemical and physical processes to GOM and/or Hg(p) which are lost from the atmosphere on time scales of days to weeks due to their higher susceptibility to wet and dry deposition [Seigneur et al., 2006]. The relatively short atmospheric lifetimes for GOM and Hg(p) compared to GEM also suggest that a higher proportion of these mercury species will deposit close to their point of emission. Given the differences in atmospheric lifetimes between elemental and oxidized/particulate-bound forms of mercury, the speciation of mercury in emissions from coal-fired power plants will largely determine the impact of coal combustion on nearby communities as well as remote environments with no local anthropogenic mercury emissions, such as the Arctic [Steffen et al., 2008].

[3] Modeling the atmospheric mercury cycle relies on global mercury emissions inventories that are in turn based on an average mercury speciation from coal-fired power plants of 50% GEM, 40% GOM, and 10% Hg(p) [Pacyna et al., 2006]. This speciation can vary considerably based on the blend of coal that is burnt, the conditions during coal firing, and on the emissions controls installed [United States Environmental Protection Agency, 1998; Edgerton et al., 2006]. Regardless, coal-fired power plants emit mercury that has the potential to impact both the local and global mercury cycling.

[4] Recent field studies indicate that mercury in precipitation downwind of North American power plants is not elevated [Seigneur et al., 2003; Prestbo and Gay, 2009], at odds with the average quantity of oxidized (i.e., “soluble”) mercury potentially emitted from these plants based on global emission inventories. Recent work by Kos et al. [2012] found that root mean square errors and biases in modeled local mercury deposition near coal-fired power plants decreased significantly when the emitted mercury speciation was shifted from the average speciation of 50:40:10 to 90:8:2 GEM:GOM:Hg(P). Plume-dilution studies [Prestbo et al., 2005] have suggested that GOM transforms to GEM in power plant plumes, presumably through reduction by another plume constituent, although contemporaneous decreases in GOM with increasing GEM were not always observed [Prestbo et al., 2000]. A multiyear ground-level monitoring campaign of mercury speciation in coal-fired plant plumes in the southern United States found discrepancies between the observed in-plume mercury speciation and the estimated emissions from nearby coal-fired power plants [Edgerton et al., 2006] based on data collected by the United States Environmental Protection Agency for U.S. coal-fired power plants. While the observations of Edgerton et al. [2006] support the hypothesis that reduction of oxidized mercury occurs in coal-fired power plant emissions, the authors of the study noted that interpretation of their results was complicated by potential losses of GOM to surface deposition and by errors associated with using nationwide emissions estimates for the individual power plants studied. A stronger case for in-plume transformation of mercury could be made if a discrepancy between in-stack and in-plume mercury concentrations was observed at other coal-fired power plants.

[5] We present here GEM, GOM, and Hg(P) measurements taken in the plume of the Ontario Power Generation's Nanticoke power station in Southern Ontario in 2000 as part of the Health Canada Toxic Substances Research Initiative (TSRI) Metals in the Environment Research Network (MITE-RN) study. Measurements were collected onboard the National Research Council of Canada's DHC-6 Twin Otter research aircraft. Measuring mercury speciation in-plume by aircraft avoids near-surface influences on mercury chemistry, such as poorly defined emissions from sources other than the plume under investigation and dry deposition to the Earth's surface. We compare measured in-stack and in-plume mercury speciation to address the question of mercury reduction chemistry in the plume and the inherent challenges of studying mercury emissions from coal-fired power plants. We discuss the role that plume mercury speciation plays in the fate of mercury emissions from Nanticoke, with specific focus on the balance between local mercury deposition and transport into the global atmosphere.

2 Site Description

[6] The Ontario Power Generation (OPG) Nanticoke Generating Station (NGS) is an eight-unit coal-fired power plant located on the northern shore of Lake Erie (Figure 1), in Southern Ontario (42.80°N, 80.52°W). All units are 500 MW Babcock and Wilcox boilers, originally placed in service between 1973 and 1978. Each boiler is supplied pulverized coal from five mills fed by dedicated silos. Boilers are fitted with Deutsche Babcock low NOx burners and cold-side electrostatic precipitators for particle collection. The coal blend in use during the TSRI MITE-RN study was a 50:50 blend of bituminous coal from the eastern United States and sub-bituminous coal (Powder River Basin (PRB) coal) from the western United States. The eight units feed into two stacks of four units each that emit at 198 m above ground level, higher than any neighboring industries, such as the Imperial Oil refinery (one multiflue 100 m stack) or U.S. Steel Works (emissions below 100 m) several kilometers to the northeast and northwest, respectively [Sahota et al., 1985].

Figure 1.

A map of relative locations of potential sources of mercury to the regional atmosphere surrounding the Nanticoke Generating Station. Sources of mercury not shown on this map are the emissions from coal-fired power plants in the Ohio River Valley, to the south-southwest of this map, as well as the Imperial Oil refinery and U.S. Steel Works in the immediate proximity of the NGS (see Figure 2).

[7] The NGS is located in a forested area with some agricultural crop lands. The largest town in the near vicinity is Nanticoke, Ontario, 19 km to the northwest. The nearest cities north of Lake Erie are London, Ontario, at roughly 100 km to the west-northwest; Hamilton and the greater Toronto urban area at roughly 50–100 km to the north-northeast; and the Niagara Falls/Buffalo region at roughly 100 km to the east-northeast. The closest city on the south shore of Lake Erie is Erie, Pennsylvania, roughly 75 km south of the study area. The nearest coal-fired plant in Canada is the Lambton Generating Station, approximately 200 km to the west of the NGS. While there are several high-output coal plants in western New York, the majority of coal-fired power plants on the U.S. side of Lake Erie lie in the Ohio River Valley, several hundred kilometers to the south-southwest of the study area. The Ohio River Valley contains a high concentration of coal-fired power plants and is one of the major sources of mercury emissions in the United States [Center for Environmental Cooperation, 2004].

[8] Mercury emissions from Nanticoke in 2000 were reported as 229 kg Hg, which comprised 11% of Hg emissions in Canada from electricity production and 2.8% of total reported anthropogenic Hg emissions from Canada [National Pollution Release Inventory (NPRI), 2000]. For reference, the Imperial Oil refinery and U.S. Steel Works near Nanticoke reported year 2000 mercury emissions of 0.95 and 21.5 kg, respectively. Three steel works located in Hamilton, to the northeast of Nanticoke, reported combined emissions of 108 kg Hg in 2000 [NPRI, 2000]. More remote emitters such as those found in the Ohio River Valley may also significantly impact the regional background of atmospheric mercury at the study site.

[9] The Ontario Power Generation occasionally collects triplicate mercury speciation measurements in Nanticoke flue gases using the Ontario Hydro Method (OHM). The closest three OHM measurements in time to this study were taken in 1999 in flue gases downstream of the electrostatic precipitator in Unit 6. The average proportions of mercury species observed were 53% Hg0, 43% Hg(II) and 4% Hg(P) by mass [Lyng et al., 2005], values which are very similar to the average mercury partitioning for North American coal-fired plant emissions mentioned above [Pacyna et al., 2006]. Follow-up measurements in 2004 on Unit 6 with a slightly different coal blend (32:68 bituminous:PRB) were consistent with the 1999 OHM speciation measurements, with an average mercury proportioning of 51:47:2 Hg0:Hg(II):Hg(P), suggesting that the mercury speciation of Nanticoke emissions may be relatively consistent over time. Although limited in number, triplicate mercury speciation measurements from other units at Nanticoke show higher proportions of particulate mercury and lower proportions of elemental mercury, with averages of 36:40:24 and 37:50:13 Hg0:Hg(II):Hg(P) in flue gases from Unit 2 in 1993 and Unit 3 in 2004, respectively. The extent to which emissions from Units 2, 3, and 6 reflect total emissions from the Nanticoke GS is uncertain.

3 Campaign Summary

[10] A description of the TSRI MITE-RN plume campaign can be found in Banic et al. [2006]; we present here only those details pertinent to the mercury analyses of the Nanticoke plume.

[11] During the campaign, 13 flights were made into the Nanticoke plume between 17 and 28 January 2000, and 12 flights were made between 12 and 21 September 2000. GEM measurements are available for 27–28 January and 12–21 September, while GOM and Hg(P) measurements were collected throughout all flights. The duration of each flight was approximately 2 h. The Nanticoke plume was sampled up to 40 km downwind to study plume ages of up to 1–2 h, in a variety of meteorological conditions, including summer and winter, day and night, light to strong winds, and well-mixed to stable boundary layers. Plumes were tracked using wind measurement systems developed over the past two decades for this aircraft.

[12] A typical flight began with a vertical sampling of the ambient air upwind of the Nanticoke plume. The aircraft would then track the plume out to 30 km, followed by vertical profiles and horizontal passes of the plume at ranges of 20 and 10 km, finishing with a run up the plume to within 2 km of the stack. At average winds of 10 m s−1, these distances correspond to plume aging times of roughly 45, 30, 15, and 3 min, respectively. Each plume pass was preceded and followed by measurements out past the edges of the plume to best characterize the ambient environment surrounding the plume. This characterization of the ambient air is critical because of the large dilution of the plume by ambient air after emission. The plume was identified by sharp increases in particle and sulfur dioxide concentration, as discussed in section 4.2. The aircraft typically spent between 30 and 90 s in plume during a horizontal plume pass. Flights spanned altitudes from 30 m above ground to several kilometers above ground, depending on the height of the top of the plume in each flight. Examples of typical flight tracks are shown in Figure 2.

Figure 2.

Flight plan for (left) 12 and (right) 20 September 2000, with locations highlighted black where the plume was identified based on particle number concentration and CO2, respectively. Note that longitude is presented here in degrees E. The Nanticoke Generating Station (square), Imperial Oil refinery (circle), U.S. Steel Stelco steel works (triangle), and airfield (diamond) are shown. The approximate location of the north shore of Lake Erie is indicated by dashed lines. The histogram in the upper left-hand corner of each plot indicates that the wind was predominately from the west-southwest, in accordance with the direction of plume travel. The plume is relatively coherent with sharp borders and is detected out to the furthest distance traveled from the NGS.

[13] The Twin Otter was equipped with a variety of 1 Hz resolution meteorological instrumentation including temperature, dew point, and vertical and horizontal winds [Banic et al., 2006], as well as instrumentation to measure particle concentration and size distribution and a variety of atmospheric trace species. Of direct relevance to this study are the instruments detailed in Table 1.

Table 1. Twin Otter Microphysical and Chemical Instrumentation
Number concentration of particlesTSI 7610>15 nm diameter particles 1 Hz measurementGrowth to droplets followed by light scatteringForward facing stainless steel sample inlet
Sulfur dioxideTECO analyzer1 Hz measurementPulsed UV fluorescenceRear-facing Teflon inlet line mounted 20 cm from aircraft skin drawing air through a Teflon filter
Carbon dioxideLICOR 6262 analyzer1 Hz measurementIR absorptionInstalled in nose of aircraft for September flights only
Gaseous elemental mercury (GEM)Tekran 2537 analyzerSampling time of 2.5 minAtomic fluorescence spectroscopyRear-facing Teflon inlet line mounted 20 cm from aircraft skin drawing air through two Teflon filters at inlet and instrument
Gaseous oxidized mercury (GOM)Quartz annular KCl-coated denuderKept at ambient T inlet impactor 10 L min−1 flowOne in-plume and one out-of-plume per flightRoof of the aircraft forward of the prop line, denuder inlet rear-facing mounted 30 cm from aircraft skin
Particulate-associated HgPall Flex 25-mm quartz filtersAt aircraft T 10 L min−1 flowOne per flightDrawn from Big Bag

[14] The aircraft was also equipped with a 0.5 m3 collapsible Teflon bag (the “Big Bag Sampler” or “BBS”). The bag was filled by air flowing through a forward facing inlet (8 cm by 8 cm) with electronically controlled inlet and outlet valves. The BBS is capable of being flushed and filled with outside air in around 2 s, which allows instruments to sample plume air for longer lengths of time than the aircraft speed would normally permit during one plume pass. At times throughout the campaign, the particulate number concentration, elemental mercury, and SO2 analyzers drew air samples from the BBS for comparison to the external “plume-average” analyses.

[15] In addition to the aircraft instrumentation mentioned above, a mobile field laboratory was installed several kilometers to the west of Nanticoke at Simcoe, Ontario (42.83°N, 80.30°W). The field site included a Tekran 2537A Automated Mercury Monitor for GEM measurement, KCl denuders for GOM sampling, and particle filters for Hg(P) sampling. The same manual methods for GOM and Hg(P) analyses were used at the ground site and on the aircraft. At the ground site, all denuders and filters were kept in an insulated box at 50°C during sampling. Samples were drawn from a height of 3.7 m above ground level and, based on results of concurrent measurements of SO2 and particle number concentration, collected air that was consistently free of the Nanticoke plume.

[16] Gaseous elemental mercury measurements were collected in-flight using a Tekran 2537A. The Tekran 2537A analyzer was calibrated using an internal permeation standard; manual injections of mercury-saturated air in-flight through the sampling line were periodically performed to assure sample integrity under the harsh sampling conditions. Uncertainties for GEM measurement are ±10%, based on comparison of duplicate measurements in the laboratory.

[17] Gaseous oxidized mercury was collected on the aircraft from plume air (winter and summer) and ambient air (summer only) onto KCl-coated denuders [Landis et al., 2002], with the denuder sampling ambient conditions drawing air before and after plume passes. The denuders were mounted on the outside of the aircraft forward of the propeller line with a 4 cm long, 1 cm diameter, curved inlet facing the rear at approximately 30 cm from the aircraft skin. The denuders were enclosed in a Teflon case and heated by an air flow from the interior of the aircraft. KCl denuders were analyzed at the Simcoe ground station by decomposition of collected mercury to Hg0 at 550°C under a 1.5 L min−1 zero-air flow into the Tekran 2537A analyzer. The total quantity of mercury collected as GOM, in picograms Hg, was then converted into a corresponding GOM concentration (in pg m−3) using the integrated flow of air passed through the denuder in-flight. Given the extremely low concentrations of GOM in air (on the order of 10−12 g/m3), only one in-plume sample was collected for each flight to maximize the quantity of analyte collected. Individual denuders were tested in the laboratory and were found to have recovery efficiencies for HgCl2(g) emitted from a diffusion source (a commonly used GOM proxy) in the range of 70–95%. Recent work by Lyman et al. [2010] indicates that in the presence of ozone, KCl denuders may undersample GOM by roughly 20%; the ultrahigh purity nitrogen used as a carrier gas for HgCl2 was ozone free and thus was not responsible for the varying GOM recoveries observed. The uncertainty for GOM measurement is estimated at ±40%.

[18] In contrast to GEM and GOM, particulate mercury samples were collected only from the BBS. One filter was collected for each flight, with air drawn from multiple Big Bags filled in the core of the plume. As such, particulate Hg data represent an average of discrete plume sampling at various distances from the source of emission. An aircraft study at a copper smelter in Rouyn Noranda, Quebec, using the BBS found no systematic difference in dilution-corrected external-air and BBS concentrations of particulate As, Cd, Cu, Ni, Pb, Se, and Zn [Wong et al., 2006]. We infer from the particulate metal data shown in Wong et al. [2006] that the BBS provides a reliable measurement of particulate mercury concentrations as well. At the Simcoe ground station, particulate mercury on sample filters was decomposed to elemental mercury at 900°C under a 1.5 L min−1 zero air stream for quantification with the on-site Tekran 2537A analyzer. Similar to GOM, total particulate mercury as Hg0 was converted to a Hg(P) concentration using the integrated flow through the filter in-flight. Ambient air was not sampled for Hg(p) in flight.

[19] The Ontario Power Generation performs a series of quality control tests on their OHM sampling train prior to use to ensure that their measurements accurately reflect in-stack mercury speciation. Impingers in the sampling train are analyzed individually to verify mercury collection efficiencies; typically 90% or greater of incoming oxidized and elemental mercury was trapped in the first KCl or KMnO4 impinger, respectively. Recovery efficiencies of each denuder in the sampling train are determined using spiked samples of mercury and in 1999 were 96% or higher.

4 Results and Discussion

[20] The quality-controlled GEM, GOM, and Hg(p) data can be found in the supporting information. We present here the summary and treatment of data pertinent to the discussion. All mercury concentrations are presented in units of mass per standard cubic meter (i.e., at 0°C and 1 atm).

4.1 GEM Correction for Ambient Pressure

[21] During aircraft GEM analyses, the detection cell of the Tekran 2537A unit was at reduced pressure as the cell vents directly to the ambient atmosphere of the aircraft which was not pressurized. The mass flow of the argon carrier gas entering the detector is constant, resulting in increased volumetric flow and a reduced residence time for mercury in the detection cell at lower pressures [Ebinghaus and Slemr, 2000]. The direct readings from the Tekran analyzer in-flight are thus expected to be biased low compared to comparable ground-level measurements. Banic et al. [2003] empirically determined the pressure dependency of the Tekran analyzer used in this study and found that the measured GEM concentration (“GEMmeas” in ng m−3) could be pressure corrected (to “GEMpc” in ng m−3) using the following relationship:

display math(1)

where Pd is the pressure at the detector cell outlet (expressed in atm). At the relatively low altitudes studied during the Nanticoke campaign, this correction typically resulted in adjustments of <5% to measured GEM concentrations. No pressure correction is necessary for GOM and Hg(P) concentrations, as these measurements were performed post flight at ground level.

4.2 Plume Identification

[22] To determine the concentration of elemental, oxidized, and particulate mercury in the Nanticoke plume, collected data must be separated into time periods of ambient air or plume air sampling. The Nanticoke plume could be distinguished in real time from the surrounding ambient air by measurement of constituents elevated in the plume with fast-response instruments, such as particle number concentration or relative CO2 concentration. For most flights, CO2 concentrations were used to identify the Nanticoke plume; for flights when CO2 was not available, the plume was identified by enhanced particulate concentration. The threshold for detection of the plume was taken as the average of ambient concentrations prior to and after each plume pass (typically two to five measurements in total) plus 2 times the standard deviation of this mean ambient CO2 concentration. Plumes identified by CO2 data were verified using a combination of SO2 concentrations, particulate concentrations, and/or wind direction as well as the location of the aircraft with respect to the stacks. For example, the plume identification for 12 and 20 September is shown in Figure 2. During the January flights and early flights in September, carbon dioxide data were unavailable, and the plume was identified using particulate concentrations (particle diameter >15 nm). An example of the particulate time series, including BBS air analyses, during a flight can be found in Figure 3 of Banic et al. [2006]. Plume identification by particulates was complicated by the presence of nearby particulate sources, in particular the U.S. Steel Lake Erie Works to the immediate northwest, but was still possible through the use of the supporting information mentioned above.

4.3 Ambient Mercury Concentrations

4.3.1 Gaseous Elemental Mercury

[23] Gaseous elemental mercury concentrations in the ambient air surrounding the NGS, shown in Figure 3, are in good agreement with other measurements of GEM in southern Canada [Blanchard et al., 2002; Banic et al., 2003] as well as in other regions of the Northern Hemisphere [Ebinghaus and Slemr, 2000; Swartzendruber et al., 2008; Ebinghaus et al., 2009]. Ambient GEM concentrations are relatively constant with height, being distributed around mean concentrations of 1.7 ± 0.2 ng Hg m−3 in September 2000 and 2.4 ng ± 0.2 ng Hg m−3 in January 2000. The frequency distributions of ambient GEM in this study are similar to those obtained by Banic et al. [2003] in central Ontario and southern Québec.

Figure 3.

Ambient gaseous elemental mercury (GEM) concentrations measured by aircraft, in ng m−3 (at STP) plotted versus (left) altitude, versus (middle) wind direction, and as (right) frequency distributions. GEM concentrations show no apparent trend with height, consistent with other observations in Canada and other regions of the Northern Hemisphere. The frequency distributions of GEM concentrations obtained in this study resemble distributions taken in central Ontario and southern Québec by Banic et al. [2003]. Air originated in this study from the west, with elevated concentrations in winter from the northwest and in summer from the southwest. Elevated ambient GEM is likely a result of local sources in winter and long-range transport in the summer.

[24] Elevated ambient GEM concentrations in the summer were observed predominately in periods of south-westerly winds. Back trajectories calculated using Canadian Meteorological Centre's trajectory model [Côté et al., 2007] for these flights indicate that air parcels arrived at Nanticoke from several hundred meters above sea level in the central United States. The probable source of elevated GEM concentrations in ambient air during September 2000 is therefore the long-range transport of emissions from coal-fired power plants in the Ohio River Valley. Winds arrived at Nanticoke from the northwest during the two flights where wintertime GEM measurements were possible. Back trajectories for these time periods suggest that winds arrive from Northern Ontario without crossing significant sources of emissions. Elevated GEM concentrations during January 2000 may instead originate from local emissions sources to the northwest of the Nanticoke Generating Station.

4.3.2 Comparison With Integrated Atmospheric Deposition Network Site at Point Petre, Ontario

[25] An additional validation of the mercury measurement techniques used in-flight can be made by comparison to long-term automated mercury measurements at an Integrated Atmospheric Deposition Network (IADN) site on the eastern end of Lake Ontario at Point Petre (43.84°N, 77.15°W). The Point Petre monitoring site is located on a small peninsula in the eastern end of Lake Ontario, roughly 260 km to the northeast of the Nanticoke station. The area surrounding the site is lightly populated and consists of mixed brush, woods, and agricultural lands. Mercury data from Point Petre are compared to in-flight mercury measurements near Nanticoke in Figure 4. Note that while GEM measurements are contemporaneous for Nanticoke and Point Petre, the closest GOM measurements in time for Point Petre were from September 2002. Ambient particulate mercury measured at the ground site upwind of Nanticoke is also shown for comparison.

Figure 4.

Comparison of mercury measurements (in-flight for GEM and GOM, at ground site for Hg(P)) in ambient air near the Nanticoke GS (black circles) and at the IADN site at Point Petre, Ontario (gray dots). Point Petre GEM measurements are from September 2000, while GOM and Hg(P) from Point Petre are from September 2002. The agreement between data sets for both sites indicates that mercury measurements at Nanticoke should fit well into both the regional and temporal context of southern Canada, and in turn, that they reliably describe mercury concentrations aloft under in-flight environmental conditions.

[26] There is very good agreement between mercury measurements at Nanticoke and at Point Petre, suggesting that mercury measurements taken at ambient temperatures and pressures aboard the aircraft do provide an accurate picture of mercury concentrations aloft in the region surrounding the Nanticoke GS.

4.3.3 Gaseous Oxidized Mercury

[27] Ambient gaseous oxidized mercury concentrations measured by the aircraft and at the ground site during flights in September 2000 are presented in Figure 5. Ambient air was not sampled by denuders during the flights in January 2000. The good agreement between the two sets of measurements suggests that the aircraft measurements are reliably recording the GOM concentration in air surrounding the Nanticoke station, despite the atypical conditions for sampling aboard the aircraft and analysis using a manual method. Although the data are sparse and show considerably variability, the average concentrations at ground level (6 ± 7 pg m−3) and at height (9 ± 7 pg m−3) are consistent with the previously reported vertical profiles of GOM in the eastern North America and Pacific [Sillman et al., 2007; Swartzendruber et al., 2009].

Figure 5.

Ambient gaseous oxidized mercury (GOM) concentrations plotted versus height. The ambient GOM concentrations at height (open circles) and ground level (closed circles) are roughly equal, with mean values of 6–9 pg m−3.

4.4 Plume Mercury Concentrations

4.4.1 Plume GEM Concentrations

[28] The Tekran automated mercury analyzer had a sampling time of 150 s, longer than the time spent in the plume for most plume passes, resulting in the collection of a mixture of plume air and ambient air during sampling. If the flow rate through the Tekran analyzer was assumed to be constant, then the plume GEM concentration (GEMP) can be estimated from measured GEM concentrations (GEMM) and ambient GEM concentrations (GEMA) using a two end-member mixing model:

display math(2)

where f is the ratio of time spent in-plume to the total time spent sampling (i.e., f = tP/150, 0 < tp ≤ 150). Both GEMA and GEMM are pressure corrected. The sharp contrast between the plume and surrounding air allowed for the estimation of tP typically to within several seconds using CO2 and particulate concentrations.

[29] For the Nanticoke summer and winter campaigns, GEMA was usually taken as the mean of ambient samplings before and after each plume pass. The only notable exceptions were instances when plume samplings were performed at altitudes significantly different than the surrounding ambient samplings. In this case, GEMA was determined from the mean of same-day ambient measurements taken at altitudes close to the plume pass in question. The resulting plume concentrations are plotted alongside ambient GEM concentrations in Figure 6. Plume GEM concentrations are elevated compared to ambient GEM concentrations, but typically by less than 2 times the ambient, consistent with rapid plume dilution after emission.

Figure 6.

Plume GEM and ambient GEM concentrations for winter and summer campaigns at Nanticoke.

[30] Once the GEM concentration in-plume for a particular pass is known, the GEM enhancement due to the plume (GEMex) may be calculated from

display math(3)

Plume-added GEMex concentrations, ranged from 6 pg m−3 to 6.4 ng m−3, with an average value for plume GEMex of 0.83 ng m−3. Plume GEM enhancements are plotted versus the age of the plume by flight in Figure 7. For the sake of clarity, error bars are omitted from Figure 7; the uncertainty in GEMex is on the order of ±30% based on propagation of errors in GEMP and GEMA. Time since emission is determined using the location of the aircraft during sampling, with respect to the Nanticoke stacks, and the average wind speed during sampling.

Figure 7.

Plume GEM enhancements, in ng m−3, plotted versus time since emission, for each day that elemental mercury was measured. Day of year is indicated in the upper right-hand corner of each graph. An “N” denotes a night flight in darkness. In most cases, plume-added GEM shows no clear trend with plume aging, within the significant variability inherent to the plume.

[31] Plume GEM enhancements show considerable variability and no general trend with plume age. Figure 7 serves as a caution against the over-interpretation of limited mercury measurements in a dynamic environment such as the Nanticoke plume. In some cases, such as day 28 or day 258, observed plume GEM concentrations seem to indicate loss or production of mercury in the plume, although comparison to data from other flights shows that these implied trends are within the inherent variability of the plume.

[32] The lack of clear trend in plume GEM with plume age implies that either for the plume ages studied in this campaign, further plume dilution is negligible and any plume mercury chemistry is unobservable, or that plume mercury dilution is balanced by in-plume production of elemental mercury. There are no clear differences between winter and summer flights, or between night and day flights. Given the lack of observable diurnal or seasonal dependencies on plume mercury concentrations, it is more likely that plume dilution is minimal after an initial rapid dilution and that any mercury chemistry occurring is hidden within the inherent variability of the plume (generally on the order of 0.5–2 ng m−3). This observation does not negate the possibility for in-plume mercury chemistry on longer time scales than those studied in this campaign, or the possibility of rapid transformation of mercury after emission [Edgerton et al., 2006; Laudal and Levin, 2006].

4.4.2 Plume GOM Concentrations

[33] Measured in-plume (P), ground-level (G), and ambient (A) GOM concentrations are presented in the left column of Figure 8. In-plume GOM concentrations tend to be significantly elevated above concentrations in the surrounding atmosphere at ground level and aloft. The GOM enhancement added by the plume to ambient air can be determined using an equation analogous to equation (3) mentioned above. In the winter, where in-flight ambient GOM measurements were not taken, ambient GOM was assumed to be equal to the same-day ground-site GOM concentration. Plume-added GOM concentrations for January and September 2000 are plotted against plume age in the right column of Figure 8. The plume age for a GOM measurement was estimated as the average of plume ages during plume passes where denuder samples were collected, weighted by the volume of air passed through the denuder during each plume pass. Plume-added GOM concentrations ranged from 6 to 530 pg m−3, with an average value of 130 pg m−3.

Figure 8.

(left column) Measured in-plume (P), ground-level (G), and ambient aloft (A) gaseous oxidized mercury (GOM) concentrations versus day of year and (right column) plume-added GOM concentrations (GOMP − GOMG; GOMP − GOMA) versus time since emission. Measured GOMP concentrations are significantly enhanced over GOMG (days 22–28) and GOMA (days 255–264) concentrations. Similar to GEM, no trend in plume-added GOM concentrations is apparent with plume aging, at least within the uncertainty of measurement (±50%).

[34] As was the case for GEM, there are no observable trends in plume-added GOM with plume aging, although the uncertainty in measured plume-added GOM concentrations is considerably larger (±50%). Considering that each GOM measurement was collected on a different flight, under different meteorological conditions, a lack of trend with plume age may not be surprising. For example, in-plume GOM concentrations above 100 pg m−3 may reflect days where the plume maintains its integrity further downwind from the plant than usual, rather than in-plume chemical production.

4.4.3 Plume Hg(P) Concentrations

[35] In-plume and ground-level particulate-associated mercury concentrations are presented in the left column of Figure 9. No ambient Hg(P) measurements were made in-flight. Plume-added Hg(P) concentrations were therefore calculated analogous to equation (3) but assuming ambient Hg(P) at height was equivalent to Hg(P) concentrations measured at the Simcoe ground site. At times, the ground-level Hg(P) concentrations were elevated and of similar value to in-plume Hg(P) (Figure 9) and likely reflect the presence of nearby transient emissions at the surface. To limit potential bias that ground-level sources might have on the calculation of plume-added Hg(P) concentrations, a median ground-level value of 2.9 pg Hg m−3 in winter and 5.4 pg Hg m−3 in summer was used to estimate ambient particulate mercury concentrations aloft. The resulting plume-added Hg(P) concentrations are plotted versus plume age in the right column of Figure 9. Plume-added Hg(P) concentrations ranged from 1 to 250 pg m−3, with an average value of 50 pg m−3.

Figure 9.

(left column) Measured in-plume and ground-level particulate-associated mercury (Hg(P)) concentrations versus day of year and (right column) plume-added Hg(P) concentrations versus time since emission. In most cases, in-plume Hg(P) concentrations are enhanced slightly above nearby ground-level Hg(P) concentrations, although summertime in-plume Hg(P) concentrations can reach several hundreds of picograms per cubic meter. There are no apparent trends in plume-added Hg(P) concentrations with aging of the plume, within the uncertainty of measurement (±40%).

[36] No clear trends are observable for plume-added Hg(P) within the measurement uncertainty (±40%). High concentrations (>100 pg m−3) were limited to the summer campaign and were not coincident with elevated GOM concentrations during the same time period.

4.5 Estimating Stack Mercury Concentrations From In-Plume Hg and SO2 Measurements

[37] The in-stack concentration of a mercury reservoir (XST) can be estimated from the respective in-plume excess mercury concentration (Xex) by

display math(4)

where fd is a dilution factor determined using in-plume and in-stack concentrations of a plume constituent that is effectively inert on the time scales of this study. Equation (4) assumes that chemical transformation of mercury does not occur in the emitted plume; if chemical reactions involving mercury do occur, then the XST estimated using equation (4) may significantly differ from measured in-stack mercury concentrations.

[38] We estimate plume dilution using plume and stack sulfur dioxide concentrations, given the comparatively large concentrations emitted by the NGS (~420 ppm SO2) and the relatively slow rate at which SO2 is lost from the Nanticoke plume, ~1–4% per hour [Anlauf et al., 1982]. Stack SO2 concentrations with a time resolution of 1 min measured using a continuous emissions monitoring system for January and September 2000 were provided by Ontario Power Generation. Plume SO2 measurements could be paired with stack SO2 concentrations based on estimated plume ages and the time of sampling. However, due to the strong gradient in SO2 on very short time scales, the aircraft-measured plume SO2 concentrations as read by the TECO analyzer were underestimated due to an incomplete sampling of the plume by the SO2 analyzer (discussed in the following section).

4.5.1 Correction for Incomplete Detector Equilibration When Sampling Strong Gradients

[39] The response time of the TECO SO2 analyzer is on the order of 100 s, determined by the time required to flush the detection cell of the instrument with sample. For comparison, the median time spent in-plume on a single pass was 60 s (n = 210), with a median absolute deviation (MAD, representing the inner 50% of the data) of 30 s. In most cases, the SO2 instrument therefore failed to equilibrate with the plume air during a pass, resulting in a lower apparent SO2 concentration than was actually present (Figure 10). Sampling air collected by the Big Bag Sampler allowed for sufficient time to accurately measure SO2 concentrations in-plume. Unfortunately, not every plume pass contained a Big Bag sampling. A method to recover plume SO2 concentrations from partial measurements during a plume pass was required to expand the available SO2 data to cover all plume passes possible.

Figure 10.

SO2 measured during plume passes is significantly lower than SO2 measured from the Big Bag Sample collected during the same pass (upper right-hand corner). There is insufficient time during a plume pass for equilibration of the SO2 detector cell with plume air, resulting in an underestimation of SO2 in the plume. Plotting the slope of the instrument response during a plume pass versus the peak SO2 measured off the following bag gives the empirical relation plotted on the left. This relation can be used to estimate SO2 for plume passes where no BB sample was collected. Estimated SO2 concentrations are significantly higher than SO2 abundance directly measured during plume passes where the BBS was filled (lower right-hand corner).

[40] A comparison of the response of the SO2 analyzer during a plume pass containing a Big Bag fill to the measured SO2 concentrations in air drawn from the Big Bag indicates that the slope of the SO2 analyzer's response during a plume pass is proportional to the SO2 concentration in-plume (Figure 10). From this empirical relation, the slope of the SO2 analyzer response during a plume pass without a Big Bag fill can be used to determine the in-plume concentration of SO2, to within ±20%. The plume SO2 concentration estimated in this manner is typically around twice the concentration directly measured from the plume.

[41] SO2 concentrations are typically in the hundreds of parts per billion range, corresponding to plume dilutions of roughly 1:1000. For comparison, the SO2 instrument detects a background SO2 concentration of 0.1–3 ppb SO2 in ambient air. SO2 concentrations in this study are roughly an order of magnitude higher than those in Edgerton et al. [2006] and reflect the ability of aircraft to study more concentrated plumes than those that may arrive at ground-level monitoring stations.

[42] Plume SO2 could be directly paired with 129 of the 162 plume GEM measurements. The remaining 33 plume GEM measurements occurred as the plane exited the plume and the above method cannot be applied to estimate plume SO2 concentrations. For GOM and Hg(P) measurements, plume SO2 concentrations were estimated by weighting individual SO2 concentrations by the total volume of air passed through the denuder or filter during the SO2 measurement. Dilution factors estimated from SO2 concentrations and equation (4) range from 2 × 10−5 to 3 × 10−3, with a median of 5 × 10−4 (n = 174, MAD of 3 × 10−4).

4.5.2 Anomalous GEMST at Low Plume SO2 Concentrations

[43] In-stack GEM concentrations (GEMST) at the point of emission, estimated from in-plume GEM concentrations and accounting for plume dilution, are plotted as a function of plume SO2 concentrations in Figure 11. The uncertainty of GEMST values is estimated at ±40%, based on propagation of errors in GEMex and SO2 concentrations. For the most concentrated plumes encountered in this study, GEMST varies little as the plume dilutes, with values near the median GEMST of 1.0 µg Hg0 m−3. GEMST appears to increase as the plume dilutes to near-ambient SO2 concentrations. When the plume is very dilute, SO2-derived dilution factors (fd) exhibit a large influence on GEMST concentrations estimated from in-plume GEM, with small variations in the in-plume GEM concentrations resulting in relatively large changes in the corresponding stack GEMST concentration. In this manner, a relatively minor source of mercury to the regional air or the entrainment of small amounts of polluted air from below the boundary layer may significantly bias GEMST. For example, at plume SO2 concentrations of less than 100 ppb, the addition of relatively small amounts of mercury (<10% of observed GEMP, or <0.2 ng Hg m−3 on average) can bias estimated GEMST by up to several µg Hg0 m−3. In extremely dilute plumes (close to ambient SO2, on the order of 1–3 ppb), this bias can easily account for the large GEMST estimates observed in Figure 11. For this reason, we exclude mercury data when SO2 is less than 100 ppb from the following discussion. SO2 concentrations estimated for GOM and Hg(P) were consistently above 100 ppb, and therefore all GOM and Hg(P) data are considered.

Figure 11.

Estimated stack GEM (GEMST) concentrations (in µg m−3) plotted against SO2 concentrations in the plume. Note that the abscissa is reversed, reflecting a dilution of the plume traveling from left to right. Elevated GEMST concentrations at low SO2 are likely the result of addition of mercury from minor sources to the dilute plume.

[44] We also note that enhanced GEMST values at SO2 concentrations above 100 ppb occur solely when the winds arrive from the northwest. Elevated GEMST concentrations from the northwest may be due to addition of mercury from other local emitters such as the U.S. Steel Lake Erie Works to the northwest of the Nanticoke plant. Any potential bias from these anomalous data points can be limited through the use of median GEMST values rather than mean concentrations.

4.5.3 Comparing Estimated and Measured Stack Mercury Concentrations

[45] In-stack GEM, GOM, and Hg(P) concentrations estimated from in-plume excesses, accounting for plume dilution, are plotted alongside reported in-stack mercury concentrations from 1999 in Figure 12. Edges of the boxes in Figure 12 represent the 25th and 75th percentiles, while whiskers extend to the 99.3% confidence interval (±2.7 sigma). Median in-stack concentrations of mercury in GEM, GOM, and Hg(P) reservoirs are given in Table 2. Note that in spite of considerable variability in plume mercury and uncertainty in mercury measurement, there is significant agreement between total in-stack mercury estimated from plume measurements and the total in-stack mercury measured by Ontario Power Generation at Nanticoke (one-sided t test of equivalence with unequal variance, p > 99%, υ = 7).

Figure 12.

In-stack mercury concentrations (XST) as estimated from in-plume mercury concentrations (XP) and as reported by Ontario Power Generation (XS). Note that Xs values have been adjusted from reported values at a reference temperature (i.e., µg Hg m−3) to concentrations at STP. Boxes contain the middle 50% of the data distributed around the median (red line in box), while whiskers show all data lying within the 99.3% confidence interval (±2.7 sigma). Crosses are values that are statistical outliers. In-stack concentrations of GEM and Hg(P) estimated from in-plume measurements agree well with reported GEM and Hg(P) concentrations in-stack, while in-plume GOM concentrations appear to be significantly lower.

Table 2. Median In-Stack Mercury Concentrations (XST), Estimated From Plume Excesses (In-Plume) and Directly Measured (In-Stack)a
µg Hg m−3 STP
  1. a

    Total mercury (THgST) is determined from the median concentrations of GEM, GOM, and Hg(P).

  2. b

    MAD, median absolute deviation, representing the inner 50% of the data.


[46] If no physical or chemical change in mercury speciation has occurred after emission, then estimated stack concentrations for the bulk Hg reservoirs, derived from in-plume observations, should be roughly constant with plume aging and should equal the concentrations observed in the stacks (XST). The distributions of GEMST derived from in-plume measurements and from in-stack measurements are not significantly different (Mann-Whitney U test, np = 126, ns = 3, p = 0.80) suggesting that the observed GEM is relatively constant after emission from the NGS. Despite a smaller sample size, particulate-associated mercury also appears to be conserved after emission (Mann-Whitney U test, np = 17, ns = 3, p = 0.92). However, the discrepancy in GOMST seen in Figure 12 is statistically significant, with rejection at the 5% significance level of the null hypothesis that GOMST,P and GOMST,S are derived from the same distribution (Mann-Whitney U test, np = 18, ns = 3, p = 0.02).

[47] Lohman et al. [2006] proposed that the apparent loss of in-plume GOM after emission [Edgerton et al., 2006] is a result of chemical reduction of GOM to GEM. Although we find a discrepancy in plume and stack GOM concentrations similar to previous observations [Edgerton et al., 2006], we find no clear evidence for an increase in plume GEM concentrations after emission. We discuss possible origins for the disagreement between stack and plume GOM measurements in the following section.

4.6 Potential Origins of Low In-Plume GOM Concentrations

[48] Edgerton et al. [2006] propose several possible mechanisms for the observed decrease in GOM concentrations in a power plant plume after emission: (a) loss of GOM to wet/dry deposition, (b) chemical conversion of GOM to GEM during transport, (c) in-plume measurement error, and/or (d) errors in emissions estimates/measurement. At the time, they were unable to definitively attribute the discrepancy to one source, although they discounted loss to deposition and favored chemical transformation of mercury, with a potential reductant being SO2. Lohman et al. [2006] modeled in-plume chemistry based on the data of Edgerton et al. [2006] and proposed that known plume chemistry could not account for observed mercury speciation. The authors tested the addition of a first-order decay in GOM with time, as well as an assumed reaction with SO2 (see section 4.6.2) and determined that both equally improved the comparison between modeled and observed in-plume GOM concentrations.

4.6.1 Wet and Dry Deposition and Aerosol Scavenging

[49] Flue gases are emitted by the Nanticoke Generating Station above the internal boundary layer established by onshore flow from Lake Erie [Kerman et al., 1982]. As it travels inland, the Nanticoke plume may fumigate to ground level if its base encounters the internal boundary layer, whose height increases inland with distance from the coast. The base of the plume could be identified for all days except 259, at heights of approximately 400–850 m above ground level, based on particulate and carbon dioxide measurements. Hence, during most flights, the plume was isolated from the land surface prior to measurement by the aircraft, and thus dry deposition to the surface is not responsible for GOM loss prior to in-plume measurement. No precipitation events occurred during flights in this study and thus wet deposition can be discounted as a GOM removal process as well.

[50] There is a significant abundance of aerosol surfaces present in the Nanticoke plume, ~0.03 cm2 cm−3 (based on data from Banic et al. [2006]), and GOM concentrations might therefore be diminished in part due to repartitioning to particulate mercury. As noted previously, there is no evidence, within the uncertainty of measurement, for an enhancement of particulate mercury concentrations with respect to stack emissions, suggesting that there is no net transfer of GOM to particle surfaces in the plume.

[51] Rutter and Schauer [2007a] found equilibration between particulate-bound and gas-phase GOM to be very rapid (on the order of seconds), with equilibrium GOM gas-particle partitioning coefficients of 1–10 m3 µg−1 for sulfate aerosols. Using a sulfate aerosol concentration of roughly 9 µg m−3 for the Nanticoke plume [Anlauf et al., 1982], the above partitioning coefficient gives an estimated equilibrium Hg(p)/GOM ratio of 0.1–1.1 [Rutter and Schauer, 2007a]. Winter Hg(P)ST/GOMST ratios in the plume range from 0.03 to 0.96. Summertime Hg(P)ST/GOMST ratios in the plume range from 0.03 to 1.9. Only flights on days 26, 257, 258, and 261 fall outside the expected equilibrium gas-particle partitioning ratio, with values of 0.03, 1.6, 1.9, and 0.03, respectively. GOM and Hg(P) appear to be in equilibrium for the majority of days studied here, regardless of seasonality or time of day. We note that this may not preclude GOM to Hg(P) transfer during and directly after emission of the plume, when particle mass concentration would be considerably higher, but this transfer may be limited by a reduced GOM uptake at the higher stack temperatures [Rutter and Schauer, 2007b].

4.6.2 Chemical Reduction of GOM to GEM

[52] Although we find no clear evidence for a transformation of GOM to GEM in the Nanticoke plume, we may still consider whether proposed reductants are capable of converting GOM to GEM in-plume after emission.

[53] Lohman et al. [2006] proposed that oxidized mercury was reduced by sulfur dioxide through a three-step mechanism suggested by Scott et al. [2003] and based on an IR product study of the reaction SO2(g) + HgO(s) [Zacharewksi et al., 1987]. The end-products of the reaction of HgO with SO2 were found to be HgS(s), HgSO4(s), and Hg2SO4(s) [Zacharewksi et al., 1987]. In-plume reaction of oxidized mercury by SO2 would thus convert particulate mercury (HgO(s)) into other forms of particulate mercury (e.g., HgSO4(s)) and would at most reduce Hg(II) to Hg(I).

[54] It may be possible that reduction of GOM occurs in the aqueous phase, in liquid layers on aerosol surfaces or cloud droplets. Sulfite and HO2 have been proposed as aqueous phase reductants of oxidized mercury [Munthe et al., 1991; Pehkonen and Lin, 1998]. The thermodynamic favorability of mercury reduction by HO2 has been questioned [Gårdfeldt and Jonsson, 2003]. At high SO2 concentrations in the Nanticoke plume (hundreds of parts per billion SO2), Hg(II) will react with sulfite to form a stable mercury-sulfur complex, math formula [Van Loon et al., 2001]. As the plume dilutes, and sulfite concentrations decrease, math formula should dissociate to redox-unstable Hg(SO3), which in turn decomposes to Hg0. Van Loon et al. [2001] suggest that elemental mercury produced from Hg(SO3) complexes with SO2(aq) to form Hg · SO2(aq), with an apparent solubility 3 orders of magnitude higher than Hg0 alone [Van Loon et al., 2001]. Reduction of oxidized mercury in the presence of sulfite may therefore occur in liquid layers on aerosols or in liquid water droplets in the Nanticoke plume as it dilutes into the regional atmosphere. Given the relatively high solubility of the Hg · SO2(aq) complex, we suggest that the reduced mercury likely remains in the particulate phase, although evaporation of liquid layers or droplets might release the Hg0 into the gas phase. We note that if reduction of GOM to Hg · SO2(aq) occurs in the Nanticoke plume, then the net transfer of mercury would be from gaseous oxidized mercury to particulate mercury. As we have previously noted, we find no evidence to support a net transfer of GOM to Hg(p). Photoreduction of oxidized mercury is known to occur, with a midday half-life of roughly a month [Lin and Pehkonen, 1999], and is likely not significant on the time scales of this study.

4.6.3 Stack and Plume Measurement Uncertainties

[55] One significant source of uncertainty is the limited number of mercury speciation measurements at Nanticoke and whether they represent total stack emissions during the 2000 campaign. We compare plume mercury concentrations to in-stack measurements from 1999 as they were closest in time to this study. Table 3 compares the concentrations of GEM, GOM, and Hg(P) measured in emissions from Unit 6 of the Nanticoke GS in 1999 to concentrations in emissions from Units 2, 3, and 6 from 1993 to 2004. As mentioned in section 2, the mercury speciation observed in emissions from Unit 6 in 1999 appears to change little over time and is similar to emissions from other units of the Nanticoke GS. Although the blend of coal burnt in 2000 was the same as that in 1999 (50:50 bituminous:PRB), the exact blend of coal used does not appear to greatly change the mercury speciation of emissions. Total mercury concentrations in emissions from other units in the Nanticoke GS tend to be higher than those for Unit 6, but are due chiefly to higher emissions of particulate mercury. Inclusion of all mercury measurements when estimating median in-stack mercury concentrations does not significantly change the values presented in Table 2. In this study, the mercury concentrations measured in emissions from Unit 6 in 1999 are thus considered representative of total mercury emissions from Nanticoke. The discrepancy between measured and estimated in-stack GOM concentrations may reflect the intercomparison of differing mercury speciation measurement techniques. In-stack measurements of GOM may be biased high by roughly 30% due to mercury oxidation occurring in filters in the Ontario Hydro Method sampling train [Kellie et al., 2004]. KCl denuders may undersample the amount of GOM present in the plume by around 20% in the presence of 30 ppb ozone [Lyman et al., 2010]. We note that the efficiency of collection and recovery of other GOM species than HgCl2 by denuders used in this study is unknown.

Table 3. Comparison of Mercury Speciation Tests at Nanticoke From 1993 to 2004ab
YearUnitBituminous:PRB BlendTHgGEMGOMHg(P)
(µg m−3)
  1. a

    Data taken from Curtis and Sills [1999] and Lyng et al. [2005].

  2. b

    PRB, Powder River Basin.

  3. c

    (Sub)bituminous coal in 1993 sourced from western Canada.

  4. d

    Reported values are averages for triplicate measurements.


[56] If we assume both ambient and plume GOM measurements were biased during the 2000 aircraft campaign, the difference between measured in-stack GOM and estimated in-stack GOM from plume measurements decreases from 0.64 µg m−3 to 0.39 µg m−3. The probability of equivalence between bias-corrected measured and estimated stack GOM concentrations remains low (Mann-Whitney U test, n1 = 18, n2 = 3, p = 0.12) but cannot be rejected on a statistical basis. There is thus a small chance that the two data sets are equivalent and that there is no discrepancy in stack and plume-derived GOMST concentrations. Further testing of existing methods for measurement of in-stack and in-plume mercury speciation and development of new methods for mercury measurement would be a great aid in addressing whether a discrepancy between stack and plume mercury speciation exists.

4.7 Mercury Speciation and The Local and Global Impact of the Nanticoke Generating Station

[57] To demonstrate the role that mercury speciation in the Nanticoke plume can have on the mercury loading to nearby Lake Erie, a modeling study was conducted with a 3-D Eulerian mesoscale meteorological boundary layer forecast model coupled with a set of air pollution transport, dispersion and deposition modules (i.e., the BLFMAPS model). Further details can be found in Appendix A. Results for two scenarios are presented here, with mercury emissions distributed between GEM, GOM, and Hg(P) according to the observed stack concentrations (case A) and observed in-plume concentrations (case B). The model did not change mercury speciation during dilution and transport of the plume. The model was run for the meteorological conditions of April and May 2005 with a total assumed mercury emission rate of 5 mg s−1. Table 4 presents results for mercury deposition to Lake Erie by mercury species and deposition process. The results of Table 4 indicate that the speciation distribution of case A results in approximately three times as much deposition to the lake than the speciation of case B. An analysis of near-field mercury deposition for a 100 km radius circle surrounding Nanticoke was also run using the case B (i.e., “in-plume” mercury speciation) emissions scenario. The model output indicated that 2% of GEM, 27% of GOM, and 24% of particulate mercury emitted by the Nanticoke GS deposited within this area. Thus, most of the emitted mercury was advected out of the study area.

Table 4. Modeled Deposition of Mercury Emitted by the Nanticoke Generating Station to Lake Eriea
April–May 2005GEMGOMHg(p)Total Hg(P)THg% of Hg Emitted
>10 µm2.5–10 µm<2.5 µm
Mass of Hg Deposited (g)
  1. a

    Total mercury emissions of 26,352 g Hg in 61 days (April–May, 2005). Mercury speciation of emissions is 45:45:10 and 90:5:5 for Cases A and B, respectively. THg = GEM + GOM + Hg(P).

Case A
Air-water exchange787008773.3
Case B
Air-water exchange1510001150.42

4.8 Recommendations for Future Aircraft Study of Coal-Fired Power Plant Plumes

[58] This study represents one of the earliest attempts to study mercury speciation in-plume using aircraft-based instrumentation. Many of the conclusions found herein are limited by the instrumentation used and the general challenges of aircraft-based measurements. In the current study, the sampling time required for GOM and Hg(P) measurement limited sampling to a resolution of one in-plume sample per flight, which results in a data set that is difficult to interpret with respect to mercury speciation in the plume over the time periods studied. Given that particulate mercury is typically a minor fraction of total in-plume mercury, it may be acceptable to collect a limited number of Hg(P) samples in future work. However, to answer the question of in-plume transformation of GOM to GEM, it will be critical that future studies use a faster-resolution measurement of GOM, preferably on a 2.5 min time scale to directly complement GEM measurement. One acceptable candidate is the thermal-decomposition/difference method of Swartzendruber et al. [2009], although current detection limits (80–160 pg m−3) will constrain its use to relatively concentrated plumes.

[59] In-stack measurement by the Ontario Hydro Method (OHM) may provide an accurate measure of mercury speciation but it suffers a similar limitation in temporal resolution as GOM and Hg(P) measurements used in this study. Future studies should coordinate with plant staff to deploy a continuous emissions monitor (CEM) for mercury speciation measurement during the field campaign, so that mercury speciation during emission can be directly back estimated from time-since-emission estimates. Being able to directly address temporal variability in stack mercury speciation will reduce much of the uncertainty found in this study.

[60] Another question with regard to stack mercury measurements at a given facility is whether in-stack mercury speciation varies significantly from unit to unit. Future studies should plan to take a number of speciation measurements for each unit throughout the campaign to ensure that mercury speciation measured by any CEM is representative of total emissions by the plant, and/or sample directly at the base of the stacks, if possible.

[61] Finally, direct chemical identification of mercury species present in the plume would greatly aid in addressing chemical transformation of mercury in the plume. Several research groups are currently developing mass spectrometric or spectroscopic instrumentation for chemical speciation of mercury in air [Deeds et al., 2009; S. Lyman and D. Jaffe, personal communication]; it will be interesting to see if these techniques can be applied to measurement in a coal-fired power plant plume.

5 Conclusion

[62] Comparison of in-stack and in-plume mercury speciation shows that plume GEM and Hg(P) concentrations can be explained by invoking only plume dilution after emission from the Nanticoke Generating Station. Plume GOM concentrations are lower than what would be expected based on dilution alone. Previous studies have observed a similar discrepancy in GOM concentrations in other coal-fired power plant plumes and have suggested that GOM may be reduced back to GEM during or after emission. However, we know of no plausible chemistry for such a process and we see no clear evidence for this transformation of mercury in the Nanticoke plume. A small chance exists that the discrepancy results solely from inherent errors in the in-stack and in-plume GOM measurement techniques.

[63] The extent to which mercury emitted from the Nanticoke GS is deposited locally depends significantly on the mercury speciation of the Nanticoke plume. Regardless, our limited modeling suggests that the majority of mercury emitted from the Nanticoke GS is transported out of the region into the global atmosphere.

[64] Further study of in-stack and in-plume mercury speciation would be of great value, given the large uncertainty that still exists regarding GOM emissions from coal-fired power plant plumes and the fate of GOM after emission. Higher resolution measurements of GOM are needed to fully describe the in-stack and in-plume variability. In many ways, this paper serves as a summary of the challenges that need to be considered and addressed during the planning of future aircraft campaign to study mercury chemistry in the plume from a coal-fired power plant.

Appendix A

[65] Mercury transport and loss in the Nanticoke plume were modeled using the Environment Canada BLFMAPS model. BLFMAPS is a 3-D Eulerian mesoscale meteorological boundary layer forecast model (BLFM) coupled with a set of air pollution transport, dispersion, and deposition (APS) modules [Daggupaty et al., 1994, 2006, 2009]. The BLFM model utilizes 20 km resolution meteorological data from the Canadian Meteorological Centre, interpolated to the 5 km horizontal grid spacing of the Eulerian model, to predict meteorological parameters for the subsequent 12 h with a 5 minute time step over a 400 × 400 km2 area. The vertical axis is split into 10 layers (0, 1.5, 3.9, 10, 100, 350, 700, 1200, 2000, and 3000 m above ground level) following local topography. To assess the local impacts of mercury emissions from the NGS, we focused modeling efforts on a circular area of 100 km radius, corresponding to transit times of roughly 4–6 h.

[66] Atmospheric pollutant transport and dispersion were solved numerically by finite difference approximation and an operator splitting scheme, with horizontal advection terms solved using a modified Bott's scheme. The predicted meteorological variables, mixed layer depth, and turbulent parameters in 3-D space and time were used by the air pollution modules to predict hourly mercury concentrations and deposition. Land use at the ground surface was modeled using land use category data from the U.S. Geological Survey Global Land Cover data set ( Mercury concentrations in the plume are assumed to only decrease due to deposition processes and not chemical transformation. The air pollution transport, dispersion, and deposition modules were originally designed for passive pollutants and particulate matter; in this study, we have modified the APS modules to suit Hg-species specific simulations, as discussed in the following subsections.

A1 Estimated Mercury Emissions

[67] Total mercury emission from the NGS varies between 4 and 8 mg/s, based on annual emissions reported to the National Pollution Release Inventory [NPRI, 2000]. For modeling purposes, we used a total mercury emission of 5 mg/s. We followed two modeling cases with differing mercury speciation based on the observed stack (case A) and plume (case B) GOM, GEM, and Hg(P) concentrations. Particulate mercury is grouped into three size bins: large particles with diameter >10 µm, medium particles with diameter between 2.5 and 10 µm, and small particles with diameter <2.5 µm. Eighty percent of the total Hg particle mass is assumed to be in the large size bin, with 5% of the mass in the medium bin and the remaining 15% of the Hg mass in the small size bin. The aerodynamic diameters selected for each bin are 20 µm, 4 µm, and 0.25 µm, respectively.

A2 Dry Deposition

[68] The dry deposition flux of mercury species was modeled as the product of their respective concentrations at 1.5 m height and their effective dry deposition velocities (Vd, cm/s) which take into account sub-grid heterogeneous land-type effects [Ma and Daggupaty, 2000; Zhang et al., 2001, 2003].

A3 Wet Deposition

[69] Wet deposition of mercury was estimated as the product of the vertically integrated mercury concentration (C(x,y)), a normalized scavenging coefficient (Λ, s−1 mm−1 h) and the estimated precipitation rate (P, mm h−1). The summer scavenging coefficients for particulate mercury were taken as 1.4 × 10−5, 2.2 × 10−4, and 1.8 × 10−3 s−1 mm−1 h for small, medium, and large particles [Gatz, 1975; Slinn, 1977; Schwede and Paumier, 1997]. In winter, the corresponding scavenging coefficients were taken as 4.7 × 10−6, 7.3 × 10−5, and 6.0 × 10−4 s−1 mm−1 h, respectively. We assigned a scavenging coefficient of 3.0 × 10−6 s−1 mm−1 h for GEM and 6 × 10−4 s−1 mm−1 h for GOM [Berg et al., 2001; Ryaboshapko et al., 2004].

A4 Air-Water Exchange

[70] The transfer of gases across the air-water interface was modeled as the ratio of the product of the transfer velocity (Kw, cm/s) and the surface air concentration (Ci) to the Henry's law constant for a specific gas (H, unitless). Transfer velocities for GEM and GOM are taken as 0.0025 cm/s and 2 cm/s, respectively [Mason and Sullivan, 1997; Lai et al., 2007].


[71] This study was jointly funded by the project 153 of the Toxic Substances Research Initiative (managed by Health Canada and Environment Canada), the Metals in the Environment Research Network and Environment Canada (EC). This work was made possible through the expert support of the pilots (John Aitken and Robert Erdos) and staff of the National Research Council of Canada-Institute for Aerospace Research and Steve Bacic, John Deary, Heidi Krall, and Phillip Cheung of EC. We thank Ontario Power Generation (OPG) for sharing stack measurements and Rob Lyng and Leonard Terplak of OPG for their insights and discussions regarding mercury emissions from the Nanticoke Generating Station. Funding for the data analysis was provided by the Clean Air Regulatory Agenda. The manuscript significantly benefited from the comments of Mark Cohen and two anonymous reviewers.