Environ. Sci. Technol. 2006, 40, 4563-4570
Mercury Speciation in Coal-fired Power Plant Plumes Observed at Three Surface Sites in the Southeastern U.S. E R I C S . E D G E R T O N , * ,† BENJAMIN E. HARTSELL,‡ AND JOHN J. JANSEN§ Atmospheric Research & Analysis, Inc., 410 Midenhall Way, Cary, North Carolina 27513, Atmospheric Research & Analysis, Inc., 730 Avenue F, Suite 220, Plano, Texas 75074, and Southern Company, P.O. Box 2641, Birmingham Alabama 35291-8195
Elemental Hg (Hg0), reactive gaseous Hg (RGM) and fine particulate Hg (HgP) were measured intermittently at three sites in the southeastern U.S. from June 2001 through November 2004. Simultaneous measurements of SO2 and NOy were used to identify plumes from coal fired power plants (CFPPs). Emission signatures and back trajectories were used to identity specific CFPPs, and to compare observed (i.e., at the site) versus expected (i.e., at the stack) Hg speciation. Results for 41 precipitation-free plume events show that observed RGM:SO2 is substantially lower (by a factor of 2-4) than expected RGM:SO2. HgP represented 2%, or less, of total-Hg in CFPP plumes, in general agreement with emission estimates. Results for 21 events, where both RGM and Hg0 could be estimated, show that totalHg (i.e., RGM + Hg0) was essentially conserved from the point of emission to the site, and that Hg0 was the dominant form (average 84%). Emission estimates, based on coal analyses and the EPRI-ICR Hg speciation model, indicate that Hg0 should represent about 42% of Hg in the observed plumes. Possible explanations for these differences include, but are not limited to, in-plume reduction of RGM to Hg0, measurement error, errors in emission estimates, and depositional losses. Further work is needed to confirm these results and to determine if they apply to CFPPs in general, or the limited set of observed CFPPs.
Introduction Recent inventories of atmospheric Hg emissions show that anthropogenic sources are an important component of the global budget (1-3). As important as the magnitude of emissions is their chemical form. Elemental Hg (Hg0) has a long atmospheric lifetime (months) and is transported over long distances from the point of emission. Oxidized forms of mercury, collectively termed reactive gaseous Hg (RGM), are removed from the atmosphere relatively quickly (atmospheric lifetime is several days) and have a local to regional range of influence. Fine particulate Hg (HgP) is efficiently removed by precipitation, but slowly removed by dry * Corresponding author phone: 919.678.1099; fax: 919.678.1159; e-mail:
[email protected]. † Atmospheric Research & Analysis, Inc., Cary, North Carolina. ‡ Atmospheric Research & Analysis, Inc., Plano, Texas. § Southern Company, Birmingham, Alabama. 10.1021/es0515607 CCC: $33.50 Published on Web 06/29/2006
2006 American Chemical Society
deposition, and therefore, has a somewhat longer atmospheric lifetime than RGM (4, 5). Coal fired power plants (CFPPs) are recognized as a significant source of atmospheric Hg (2, 6, 7). In 1999, the U.S. Environmental Protection Agency conducted an information collection request (ICR) to assess Hg emissions from U.S. CFPPs. The ICR was implemented in three phases which documented the following: (1) coal type, burner configuration, and control technologies for all U.S. CFPPs; (2) coal composition for all U.S. CFPPs; and (3) stack emissions of total and speciated Hg at 81 U.S. CFPPs. Results of the ICR showed that CFPPs accounted for roughly one-third of anthropogenic Hg emissions in the U.S. in 1999. In addition, they showed that the form of Hg in CFPP emissions is a complex function of fuel composition, combustion characteristics, and pollution control technology. For units burning bituminous coal with cold-side electrostatic precipitators (ESPc), average Hg composition was found to be 78% RGM, 20% Hg0, and 2% HgP (8). This paper presents results from three research sites in the southeastern U.S. equipped to make high temporal resolution measurements of speciated Hg and tracer gases (i.e., SO2, NOy, and CO) for identification of sources. Continuous data are used to analyze opportunistic CFPP plumes as they are observed at each site. SO2 data are used to identify CFPPs and to calculate ratios of ambient Hg species, which are compared with emission ratios (ERs), based on day- and plant-specific emission information. Similarities and differences between observed and expected ERs are illustrated, and reasons for differences are discussed.
Experimental Section Measurements of Hg0, RGM, fine particulate Hg with aerodynamic diameter e2.5 µm (HgP), and trace gases were performed at three Southeastern Aerosol Research and Characterization (SEARCH) sites shown in Figure 1a: Yorkville, GA (YRK); Jefferson Street, GA (JST); and Outlying Landing Field #8, FL (OLF). Brief site descriptions are provided below. Additional details on SEARCH can be found in Hansen et al. (9). Yorkville (lat 33.9283 N, long. 85.0455 W) is a rural/ agricultural site 55 km WNW and 40 km SSW of Atlanta, GA and Rome, GA, respectively. The site is on a broad ridge (elevation 395 m) in a large (>150 ha) clearing devoted largely to pasture. Plume events observed at Yorkville include the following CFPPs (Figure 1b): Plant Bowen (25 km NNE), Plant Hammond (44 km NW), Plant Wansley (57 km S), and Plant Gaston (149 km SW). Hg measurements were made at Yorkville from June 2001 through November 2003. Jefferson Street (lat 33.7775 N, long. 84.4167 W) is an urban/industrial-residential site 4.5 kilometers NW of downtown Atlanta, GA. Lowery Street (3500 vehicles/day) is 60 m W of the site, and several major thoroughfares (20 00075 000 vehicles/day) are within 600-1500 m. Other sources of gaseous and particulate pollutants include a bus maintenance depot (250 m S) and a metals recycling facility (700 m E). Plant McDonough (7.4 km NW) is the only CFPP routinely observed at Jefferson Street (see Figure 1c). Hg measurements were made at Jefferson Street from January 2002 through April 2003. OLF (lat 30.5500 N, long. 87.3736 W) is a suburban site 21 km NW of Pensacola, FL and 20 km N of the Gulf of Mexico. The site is adjacent to a paved, lightly traveled (500 ha) grasscovered field. Highways I-10 and U.S. 90A are located 1.1 km N and 1.6 km S, respectively. Plant Crist is 14.4 km ENE of VOL. 40, NO. 15, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4563
FIGURE 1. Locations of Yorkville (YRK), Jefferson Street (JST), and OLF in the SEARCH network (a) and CFPPs in the vicinity of Yorkville (b), OLF (c), and Jefferson Street (d). the site (see Figure 1d), and several industrial facilities (including a paper mill) are 15-20 km N-NE. Hg measurements were made at OLF from January 2004 to the present (January-November 2004 reported here). Atmospheric Measurements. Speciated Hg measurements were made 5-6 m above ground level with a Tekran Model 2537A Hg Analyzer, a model 1135 particulate Hg unit, and a model 1130 speciation unit. The measurement approach is similar to that described by Lindberg and Stratton (10) and Landis et al. (11). Sample air was drawn at 9-10 L min-1 through a heated (50 °C) quartz impactor inlet, a KCl coated quartz annular denuder, a quartz particulate filter, then into the Hg analyzer. In sample mode, RGM was collected on the denuder and HgP was collected on the particulate filter. Hg0 was transmitted quantitatively to the Hg analyzer and trapped on a gold tube preconcentrator inside the 2537A, then thermally desorbed in argon and detected via cold vapor atomic fluorescence spectroscopy (CVAFS). In desorption mode, the particulate filter and a downstream pyrolyzer unit were heated to 800 °C. HgP was converted to Hg0, then transmitted in zero air to the 2537A for detection. Finally, the quartz denuder was heated to 500 °C to release trapped RGM as Hg0, which was then transmitted to the 2537A for detection. In mid-2003, a soda lime trap was installed in the sample line immediately upstream of the 2537A to improve the longevity and stability of gold tube preconcentrators. The Hg analyzer was calibrated approximately every 24 h with an internal permeation source of Hg0. The permeation rate of the internal source was verified against an external standard (Tekran model 2025) at least annually. We note that RGM is operationally defined, and the exact chemical composition of RGM is unknown. The KCl denuder 4564
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 15, 2006
technique has been shown to quantitatively retain HgCl2 (a likely RGM component), to quantitatively convert HgCl2 to Hg0 upon heating and, for all practical purposes, to retain no Hg0 or HgP (11). For these reasons, automated and manual KCl denuder techniques have been adopted as the method of choice for RGM measurements virtually worldwide (12-17). Independent confirmation of RGM and HgP measurements was accomplished by manual collection of air samples using a KCl impregnated quartz denuder and a quartz filter (URG, Carrboro, NC) provided by Frontier Geosciences, Inc. (Seattle, WA). The denuder was identical to that used in the Tekran, except it had a threaded outlet for connection of the quartz filter pack. Samples were collected at 10 L min-1 over 4-h periods, then shipped to Frontier Geosciences for analysis. Analysis was performed via thermal desorption followed by CVAFS. Each sample batch included four collectors, of which three were exposed and one was a field blank used for blank correction. Continuous (1-min) measurements of SO2, NOy, and CO were made at a reference height of 10 m above ground level. SO2 was measured via pulsed UV fluorescence with a TEI model 43ctl analyzer operated on a 0-200 ppb scale. The instrument was calibrated twice daily by method of additions (MOA) and zeroed 10 out of every 90 min by diverting sample air through a sodium carbonate impregnated annular denuder (URG, Carrboro, NC). CO was measured via nondispersive infrared absorption using a TEI Model 48ctl analyzer operated on a 0-3000 ppb scale (0-10 000 ppb at JST). The CO analyzer was calibrated twice daily by MOA and zeroed every 90 min by diverting sample air through a heated (50-100 °C) trap containing 200 g of 1% Pt on alumina (DeGussa, Seviersville, TN). NOy was measured
via ozone-NO chemiluminescence following reduction to NO on a 350 °C Mo converter, using a dual-channel TEI Model ctl NO-NOx analyzer operated on a 0-200 ppb scale. The NOy analyzer was zeroed four times per day and subjected to MOA calibrations twice daily. Converter performance was checked daily via MOA with n-propyl nitrate. Based on NIST guidelines (18) and a coverage factor of 2, uncertainties for Hg0, RGM, HgP, and SO2 are 12, 22, 30, and 14%, respectively. Uncertainties for the ratios Hg0:SO2, RGM:SO2, and HgP:SO2 are 18, 26, and 33%, respectively. Trajectory Calculations. Twenty-4 h back trajectories were generated using the interactive version of the NOAA HYSPLIT4 model on the NOAA-ARL web site (19). Back trajectories used EDAS 40 km meteorological data and default vertical motion, with starting heights of 1000, 500, and 250 m, for the time (hour) of peak SO2 concentration during each event. The 250 m trajectory was used to estimate transit time from point source to site, as well as time of emission at the source. CFPP Emissions. Plant-specific emissions data were obtained from Southern Company for days with plume events, or the day before, depending on estimated time of emission. Hourly SO2 and NOx emissions were measured directly using in-stack continuous emission monitors (CEMs). Coal samples (one grab sample per unit) were analyzed for total-Hg, chloride, heat content, and other characteristics necessary to calculate total and speciated Hg emissions. Hourly emissions of total-Hg, Hg0, RGM, and HgP were estimated using coal feed rates and the EPRI-ICR model equations for CFPPs which burn bituminous coal and use ESPc for particulate control. The EPRI-ICR model is a correlation model developed with ICR coal analysis data and stack test data. The uncertainty of EPRI-ICR nationwide CFPP Hg emission estimates for 1999 is on the order of 5%. The uncertainty of emissions estimates for individual CFPPs is on the order of 20-40% (7).
Results and Discussion Continuous versus Manual Measurements. Results of the continuous vs manual sampler comparisons are summarized in Table S1. For RGM, mean values for continuous and manual were 12.5 pg/m3 and 11.2 pg/m3, respectively. Average error and average % error were 1.2 pg/m3 and 2.7, respectively. The continuous analyzer tended to be lower than manual at RGM concentrations below 5 pg/m3, but higher at concentrations above 15 pg/m3 (i.e., concentrations relevant to plume analyses). For HgP, mean values for continuous and manual were 5.3 pg/m3 and 7.7 pg/m3, respectively. Average error and average % error were -2.4 pg/m3 and -30.2, respectively. Results indicate substantial underestimation of HgP by the continuous method across the range of observed concentrations. Although this finding is of some concern when it comes to determination of long-term average concentrations, it has little bearing on results reported here because HgP accounts for
