Environ. Sci. Technol. 2004, 38, 4276-4284
Total Potential Source Contribution Function Analysis of Trace Elements Determined in Aerosol Samples Collected near Lake Huron S . R . B I E G A L S K I * ,† A N D P . K . H O P K E ‡ Nuclear Engineering Teaching Laboratory, The University of Texas at Austin, R9000, Austin, Texas 78712, and Department of Chemical Engineering, Clarkson University, Potsdam, New York 13699
Aerosol samples were collected at the rural Burnt Island Ontario Integrated Air Deposition Network air sampling station on the northern shore of Lake Huron from 1992 through 1994. The samples were analyzed for trace elements by neutron activation analysis, and the air concentrations of over 30 elements were determined. Total potential source contribution function analysis (TPSCF) was used to determine the most probable geographical location of these aerosols’ origin. The TPSCF results for As, In, Sb, Se, Sn, and Zn are highlighted in this paper. Source regions for these elements ranged from Alma, Quebec, Canada to Carrollton, GA. Because of large seasonal variations in the concentrations of the atmospheric concentrations of these elements, TPSCF values were calculated for the summer and winter halves of the year as well as the entire year.
Introduction The Great Lakes Water Quality Agreement between the United States and Canada was first signed in 1972 and then renewed in 1978. This agreement outlines the desires of both the United States and Canada to restore and maintain the chemical, physical, and biological integrity of the Great Lakes Basin Ecosystem (1). Annex 15 specifically addresses the issue of airborne toxic substances and the need for monitoring these substances. Atmospheric deposition of toxic substances tends to be the dominant pollution pathway into the rural and remote regions of the Great Lakes. The research included in this paper contributes to the understanding of the sources that contribute to the toxic atmospheric deposition into rural regions of the Great Lakes. Trace elements are important for this study because they may be utilized as markers for aerosol source terms. While some of the elements in this study may be toxic to the environment, this paper focuses on geographically locating aerosol source terms and not on the environmental toxicity of aerosols. Issues regarding the magnitude of toxic element inputs into the Great Lakes are addressed in other publications (2, 3). This work compliments these previous publications and provides location information to the anthropogenic sources. Cheng et al. (4) used PSCF analysis to determine the source regions of particulate sulfate observed at Dorset, ON. Gao et
al. (5) used potential source contribution function (PSCF) analysis and Total PSCF (TPSCF) to study the source-receptor relationships for As, In, Mg, Se, and V found in precipitation and airborne particles collected at Dorset, ON. Blanchard et al. (6) reported results for total gas mercury (TGM) measurements at Burnt Island and reported on the TGM source regions in Ontario, Canada. The data included in this work were collected in the Integrated Atmospheric Deposition Network (IADN). In 1990, the Integrated Atmospheric Deposition Network Implementation Plan was developed by a Canada/U.S. Coordination Committee (7). This plan defined the time schedule, determined the chemicals to be measured, defined sites of sampling stations, set sampling protocols, and defined quality assurance and quality control (QA/QC) measures. In 1988, the Atmospheric Environment Service (AES) of Environment Canada embarked upon a program of research to address the issue of toxic deposition into the Great Lakes basin. This work was mandated under Annex 15 of the Great Lakes Water Quality Agreement. Since that time, AES has participated in IADN by setting up Canadian air monitoring stations that include the master stations of Burnt Island and Point Petre. The United States has also set up master air monitoring stations at Eagle Harbor, Sleeping Bear Dunes, and Sturgeon Point. Master stations are the primary stations set up in the IADN network, and there is one master station per lake. Ten satellite stations are also part of the IADN network. Burnt Island was chosen for this in-depth study due to its central location in the Great Lakes and its distance from urban pollution sources. While the results for TPSCF analysis are similar for Egbert and Point Petre, the Burnt Island data provides better insight into the source terms that affect remote regions of the Great Lakes.
Neutron Activation Analysis Neutron activation analysis (NAA) was chosen for this study due to the ease of sample preparation, the high degree of accuracy and precision achieved by this method, and the ability to determine over 30 elemental concentrations with a single analytical method. Aerosol samples were collected at Burnt Island for 24 h every 6 days for this study. Each sample along with a set of blank filters was analyzed for its elemental composition by NAA. Whatman 41 20 × 25 cm filters were utilized for the collection of particles at Burnt Island. Each filter was split in to eight strips (5 × 6.25 cm). Two of the 1/8 samples were used for NAA. The remaining filter material was used for other analyses and for archiving. For this work, four irradiations were performed to optimize the detection capability for the large suite of elemental concentrations to be determined. Table 1 shows the methodology used for the NAA. One of the 1/8 samples underwent the short thermal irradiation and the long thermal irradiation. The other 1/8 sample was used in the two epithermal irradiations. For three out of the four irradiations, the samples were counted in both a normal and a Compton suppressed counting mode. The Compton suppressed counting mode improved the detection limits for a number of the elements in this study. A detailed explanation of the NAA methodology and detection limits may be found in Biegalski and Landsberger (8).
Potential Source Contribution Function Analysis * Corresponding author phone: (512)232-5380; fax: (512)471-4589; e-mail:
[email protected]. † The University of Texas at Austin. ‡ Clarkson University. 4276
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 16, 2004
PSCF analysis is a conditional probability modeling technique utilized to explore the origin of chemical emissions (9, 10). The PSCF is a conditional probability that an air parcel with 10.1021/es035196s CCC: $27.50
2004 American Chemical Society Published on Web 07/09/2004
TABLE 1. NAA Methodology irradiation
decay
counting
irradiation facility neutron energy
counting mode
elements determined
60 s at 1500 kW 300 s at 500 kW 1.5 h at 1500 kW 8 h at 1500 kW
400 s 360 s 15-30 h 4-6 weeks
600 s 900 s 45 min 5h
thermal epithermal epithermal thermal
normal and Compton suppressed normal and Compton suppressed normal and Compton suppressed normal
Al, Br, Ca, Cl, Cu, K, Mn, Na, Ti, and V Cu, I, In, Si, Sn, and U As, W, Au, Sm, Br, and Sb Ag, Ce, Co, Cr, Cs, Eu, Fe, Hf, Ni, Rb, Sb, Sc, Se, Ta, Th, and Zn
a specified level of a constituent concentration arrives at a receptor site after having passed through a specific geographical area. This method combines aerosol composition data with air parcel back-trajectory calculations to determine PSCF values. The results of the PSCF calculations highlight geographic areas that are likely to have contributed to high levels of atmospheric pollutants at the receptor site. PSCF values may range from 0 to 1. A PSCF value of 0 indicates a low probability of a source being in that geographic region. A PSCF value of 1 indicates that there is a high probability of there being a source in that geographic region. Air Parcel Back-Trajectory Calculations. Air parcel backtrajectories were calculated for this work utilizing the Atmospheric Environment Service of Candada’s AES-3D trajectory model (11). Trajectories were calculated for dates from January 1, 1992, through December 31, 1994, for the Burnt Island sampling station. The backward (upwind) trajectory characterizes the path of an air parcel arriving at the sampling station. The back-trajectory path was defined at 6 h backward intervals by the latitude, longitude, and height of the point. The trajectories were calculated backward for 120 h (5 days), four times a day (0, 6, 12, and 18 UTC), at two pressure levels (1000 and 925 mbar). The pressure levels were correlated to the heights directly above the sampling station at the start of the back-trajectory calculation. The AES-3D code calculated the vertical motion of the air parcels that often resulted in cross-isobaric transport. So, the air parcels did not maintain the same height and pressure level for the duration of the back-trajectory. The results at the end point of the back-trajectory may attain appreciable uncertainty after 120 h (12). Potential Source Contribution Function Calculation. On the basis of the calculation methods detailed in Zeng and Hopke (13) and Cheng et al. (4), a computer code was written specifically for this research to calculate the PSCF and TPSCF values. This code divided the space up into one degree latitude by one degree longitude grid cells. The code read each trajectory endpoint and checked to see if there was a sampling day commensurate with the back-trajectory. The backtrajectories were calculated at 6 h intervals every day of the year, while aerosol samples were only collected for a 24 h period once every 6 days. After all the wind back-trajectories were reviewed by the code, the PSCF values for the 1000 and 925 mbar pressure levels were calculated. TPSCF values were then calculated by combining the 1000 and 925 mbar pressure level PSCF values as given in Cheng et al. (4). TPSCF values were calculated for each trace element analyzed by NAA. Cheng et al. (4) reported many of the effects of seasonality, trajectory endpoint interpolation, and selection of criterion values on the calculated PSCF values. The PSCF values were shown to change between winter and summer months due to seasonal differences in meteorology and photochemistry. The interpolation of trajectory endpoints was shown to degrade the quality of the results. The use of different criterion values (50th percentile, 75th percentile, and average value) found different results. The use of the average value as the criterion value was shown to be reasonable; however, specific cases benefited from the use of the 50th percentile value as the criterion value. It was also determined that 850 mbar
TABLE 2. Average Criterion Values Used for PSCF and TPSCF Calculations
element
yearly used as criterion value (ng/m3)
summer used as criterion value (ng/m3)
winter used as criterion value (ng/m3)
As In Sb Se Sn Zn
0.3978 0.0056 0.1289 0.6238 0.4002 6.6497
0.2971 0.0022 0.1323 0.6346 0.3511 6.4851
0.5114 0.0103 0.1257 0.6791 0.6791 6.8639
TABLE 3. Source Terms Reported at Burnt Island (14) source term
elements in source term
crustal soil/incinerator crustal salt fossil fuel burning or smelting Ni smelting
Al, Ca, Ce, Fe, Mn, Sc, Si, Ti, and V Fe, Mn, Sb, Se, and Zn Ce, Fe, K, Sc, and Zn Br, I, and Na Ag, As, and Se In
PSCF calculations moved the apparent sources too far toward the periphery. TPSCF plots were created using the data from all of the months of the year. In addition, seasonal TPSCF plots were also created. Winter TPSCF values were calculated using data from October to March. Summer TPSCF values were calculated using data from April to September. The criterion values for the seasonal TPSCF calculations were the average respective elemental concentration during those periods, whereas the criterion value for the yearly average TPSCF calculation was the average respective elemental concentration for the entire year. Therefore, the yearly average and winter and summer TPSCF plots each have different criteria values. Once the TPSCF values were calculated, a surface map of North America was made to illustrate them. Each grid space is plotted with a scaled block directly representative of the TPSCF value. The darker the grid space, the higher the TPSCF value. Table 2 provides the criteria values for the total year and summer and winter TPSCF plots for the elements discussed in this paper.
TPSCF Data Results While TPSCF plots were made for each element determined by NAA, this paper will only discuss the results for As, In, Sb, Se, Sn, and Zn. These elements were chosen because they had signals well above the NAA detection limit, the filter blank values were in control, and they were representative of the sources identified in the factor analysis described by Biegalski et al. (14). The source groups of this study as detailed in Biegalski et al. (14) are summarized in Table 3. Results and discussion for Al, Br, Cl, Fe, I, Na, and V may be found in the Supporting Information. VOL. 38, NO. 16, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4277
FIGURE 1. Arsenic TPSCF plots for Burnt Island. The plots show the TPSCF for all the data, the summer data (April to September), and the winter data (October to March). The stars represent the location of the Burnt Island station. Tin was not included in the factor analysis performed by Biegalski et al. (14) at the Burnt Island location due to the number of sampling days where the Sn level was not above the filter blank value. Tin may be utilized in this work since TPSCF only utilizes the samples with high concentrations that are significantly above the filter blank level. Biegalski et al. (14) show that for other Great Lakes stations, tin is associated with smelter sources. This source association appears to also hold true at Burnt Island. 4278
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 16, 2004
For each element, three TPSCF plots are provided along with a box and whisker plot. Yearly average and summer and winter TPSCF plots are given to show how source regions for specific elements change with the season. Dark regions of the TPSCF plots indicate that there is a high probability of there being a source in that geographic region. Light yellow regions are geographic areas that have a low probability of having a source. Points on the map that do not have a TPSCF value plotted are locations that have no wind back-trajectory
FIGURE 2. Indium TPSCF plots for Burnt Island. The plots show the TPSCF for all the data, the summer data (April to September), and the winter data (October to March). The stars represent the location of the Burnt Island station. end points in this study. The box and whisker plot shows the seasonal trend of the element’s atmospheric concentration. A line in the box represents the mode of the monthly trace element air concentrations. The range of the box depicts the bounds of the 25th and 75th percentiles of the data. This whiskers extending from the box represent the 10th and 90th percentile of the data. Data points that extend beyond the 10th and 90th percentiles are plotted individually. Each month represents the pooled data from the three years in
this study. When the box and whisker plot shows a strong seasonal variation in an element’s atmospheric concentration, the TPSCF plots show different source regions for the winter and summer. Burnt Island is located just off the southwest shore of Manitoulin Island in Lake Huron. The station elevation on Burnt Island is 185 m above sea level. A star on the TPSCF indicates the location of the station. VOL. 38, NO. 16, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4279
FIGURE 3. Antimony TPSCF plots for Burnt Island. The plots show the TPSCF for all the data, the summer data (April to September), and the winter data (October to March). The stars represent the location of the Burnt Island station. Arsenic. The factor analysis results for these data groups arsenic in a fossil fuel or incinerator source term along with Ag, As, and Se. Hoff and Brice (15) report As resulting from smelter emissions in this region. Figure 1 includes the seasonal As concentrations and TPSCF plots representing the origin of As aerosol sources to the Burnt Island area. The yearly average plot, summer plot, and winter plot each highlight different source regions. The yearly average TPSCF plot, utilizing the yearly average As concentration as the criterion value, shows that the predominant As sources come 4280
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 16, 2004
from the eastern half of the United States. Regions in the United States with especially high As potential factors occur around New York City, the southern part of West Virginia, and the region where Mississippi, Tennessee, and Alabama meet. Another strong source region appears to be in Alma, PQ. The As summer TPSCF focuses on the New York City area that was observable in the yearly average TPSCF. The As winter TPSCF highlights the New England and Carrollton, GA, area of the United States and has high probabilities in the boundary regions of northern Quebec. Gao et al. (5) also
FIGURE 4. Selenium TPSCF plots for Burnt Island. The plots show the TPSCF for all the data, the summer data (April to September), and the winter data (October to March). The stars represent the location of the Burnt Island station. saw high PSCF values in the same regions of Quebec in the examination of As at Dorsett. Indium. Indium is often associated with smelting (16) (particularly Ni smelting). Incineration is another possible anthropogenic aerosol In source (17). Figure 2 shows the seasonal concentrations and TPSCF plots for In at Burnt Island. The yearly average In TPSCF shows a high probability source location in two independent locations in Quebec. One of the regions points to the smelter facilities near Val d’Or. The second region is close to the region that Gao et al.
(5) found but were not able to identify. The Sudbury, ON, smelter is not seen due to its close proximity to the station since the emissions are released from a very tall stack. There are also source regions around New York City and in the Northwest Territories. The summer In TPSCF shows the same source regions as the yearly average TPSCF calculations. The signature from the Northwest Territories is much more evident in the summer TPSCF and is likely the result the gold mining industry near Yellowknife, NT. The winter In TPSCF is much different than the yearly average and summer TPSCF VOL. 38, NO. 16, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4281
FIGURE 5. Tin TPSCF plots for Burnt Island. The plots show the TPSCF for all the data, the summer data (April to September), and the winter data (October to March). The stars represent the location of the Burnt Island station. values. The winter In TPSCF only shows a high probability region in the area surrounding Norfolk, VA. Antimony. The Sb contained in the Burnt Island aerosol is highly enriched in comparison to what would be expected from a crustal source term. Sb is often used as a marker for incineration releases. However, it may also be associated with coal-fired power plants (18). The seasonal concentrations and TPSCF plots for Sb aerosol at Burnt Island are shown in Figure 3. These plots show that the source area of Sb is the midAtlantic United States. This source region is 4282
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 16, 2004
much more distinct in the summer TPSCF (Figure 3c) than in the winter TPSCF (Figure 3d). The Philadelphia, PA, region is of specific interest for Sb: Olmez et al. (19) report that Sb emission from a large Sb ore roaster in this region are particularly high. Selenium. The factor analysis results showed that selenium covaried strongly with Sb. Atmospheric Se is often associated with the burning of coal and fuel oils (20, 21). Figure 4 shows the Se seasonal atmospheric concentrations and TPSCF results for Burnt Island. The Se TPSCF results are
FIGURE 6. Zinc TPSCF plots for Burnt Island. The plots show the TPSCF for all the data, the summer data (April to September), and the winter data (October to March). The stars represent the location of the Burnt Island station. very similar to the Sb TPSCF results. The Se results are also similar to those found in Gao et al. (5) for Se at Dorset, ON. The source region is the midAtlantic United States. High probability regions in Pennsylvania, West Virginia, and Tennessee are commensurate with the location of many coalfired power plants in the United States. This source region becomes less distinct in the Se winter TPSCF than in the summer Se TPSCF results, presumably due to more limited northward transport in the winter.
Tin. Tin atmospheric aerosols are commonly a result of smelting. Figure 5 shows the atmospheric concentrations of Sn and the Sn TPSCF values for Burnt Island. The yearly average TPSCF figure does not show any distinct high probability regions. However, the summer TPSCF for Sn at Burnt Island has a strong source region between New York City and Philadelphia including New Jersey and Delaware. Zinc. The atmospheric aerosol concentrations of Zn at Burnt Island are highly enriched above crustal levels. The VOL. 38, NO. 16, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4283
sources of Zn to the atmosphere include incineration and smelting of nonferrous metal ores (22, 23). In Figure 6, zinc has strong seasonal fluctuations at Burnt Island making the yearly average TPSCF results less meaningful than the winter and summer TPSCF plots. The summer TPSCF plot (Figure 6c) shows elevated TPSCF values around Richmond, VA. This is similar to what was reported for Zn from Dorset, ON, by Gao et al. (5). The Richmond, VA, area does not appear to be a significant source region as illustrated in the winter Zn TPSCF calculation, but a winter region along the Georgia/ Alabama border is evident. This work utilizes TPSCF to determine the sources of atmospheric aerosols at Burnt Island on the northern part of Lake Huron. These results indicate that the primary sources for anthropogenic aerosols originate in the United States. Even though weather systems travel from west to east, the principal sources of atmospheric pollution at Burnt Island are located in the eastern and southeastern United States. Some smelting sources and urban areas in Canada were also depicted. The TPSCF results aided in the documentation of the oceanic origin of salts in the Great Lakes region. Because of the strong seasonal variation of atmospheric aerosol trace metal concentrations, it was often necessary to separate the TPSCF analyses into winter and summer months.
Acknowledgments We express our gratitude to the Canadian Atmospheric Environment Service for collecting the samples utilized in this work and funding the analysis of the samples via NAA.
Supporting Information Available Results and discussion for Al, Br, Cl, Fe, I, Na, and V. This material is available free of charge via the Internet at http:// pubs.acs.org.
Literature Cited (1) Great Lakes Water Quality Agreement, with Annexes and Terms of Reference, between the United States and Canada signed at Ottawa, November 22, 1978; International Joint Commission United States and Canada; 1989. (2) Biegalski, S. R.; Landsberger, S. Biol. Trace Element Res. 1999, 71-72, 247. (3) Biegalski, S. R. Ph.D. Dissertation, University of Illinois at Urbana-Champaign, 1996.
4284
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 16, 2004
(4) Cheng, M. D.; Hopke, P. K.; Zeng, Y. J. Geophys. Res. 1993, 98 (D9), 16839. (5) Gao, N.; Hopke, P. K.; Reid, N. W. J. Air Waste Manage. Assoc. 1996, 46 (11), 1035. (6) Blanchard, P.; Froude, F. A.; Martin, J. B.; Dryfhout-Clark, H.; Woods, J. T. Atmos. Environ. 2002, 36 (23), 3735. (7) Canada/U.S. Coordination Committee on Annex 15, Integrated Atmospheric Deposition Network Implementation Plan; Canada/ U.S. Coordination Committee on Annex 15; 1990. (8) Biegalski, S. R.; Landsberger, S. J. Radioanal. Nucl. Chem. 1995, 192 (2), 195. (9) Malm, C. W.; Hohnson, C. E.; Bresch, J. F. In Receptor Methods for Source Apportionment; Pace, T. G., Ed.; Air Pollution Control Association: Pittsburgh, 1986. (10) Lucey, D.; Hadjiiski, L.; Hopke, P. K.; Scudlark, J. R.; Church, T. Atmos. Environ. 2001, 35, 2979. (11) Olson, M. P.; Oikawa, K. K.; Macafee, A. W. A Trajectory Model Applied to the Long-Range Transport of Air Pollutants; Technical Report; Atmospheric Environment Service: Downsview, ON, Canada, 1978. (12) Kahl, J. D.; Harris, J.; Herbert, G. A.; Olson. Tellus 1989, 41B, 524. (13) Zeng, Y.; Hopke, P. K. Atmos. Environ. 1989, 23, 1499. (14) Biegalski, S. R.; Landsberger, S.; Hoff, R. M. J. Air Waste Manage. Assoc. 1998, 48, 227. (15) Hoff, R. M.; Brice, K. Proceedings of Annual Meeting of the Air and Waste Management Association; AWMA: Denver, CO, 1993; 93-RP-137.04. (16) Germani, M. S.; Small, M.; Zoller, W. H.; Moyers, J. Environ. Sci. Technol. 1981, 15 (3), 299. (17) Greenberg, R. R.; Zoller, W. H.; Gordon, G. E. Environ. Sci. Technol. 1978, 12 (5), 566. (18) Maenhaut, W.; Kauppinen, E. I.; Lind, T. M. J. Radioanal. Nucl. Chem. 1993, 167 (2), 259. (19) Olmez, I.; Sheffield, E.; Gordon, G. E.; Houck, J. E.; Pritchett, L. C.; Cooper, J. A.; Dzubay, T. G.; Benett, R. L. J. Air Pollut. Control Assoc. 1988, 38, 1392. (20) Andren, A. W.; Klein, D. H.; Talmi, Y. Environ. Sci. Technol. 1975, 9 (9), 856. (21) Winchester, J. W.; Nifong, G. D. Water, Air, Soil Pollut. 1971, 1, 50. (22) Kim, K. D.; Fergusson, J. E. Sci. Total Environ. 1994, 144, 179. (23) Lynch, A. J.; McQuaker, N. R.; Brown, D. F. J. Air Pollut. Control Assoc. 1980, 30 (3), 257.
Received for review October 27, 2003. Revised manuscript received May 21, 2004. Accepted May 21, 2004. ES035196S
