ARTICLE pubs.acs.org/est
Using Trace Elements in Particulate Matter To Identify the Sources of Semivolatile Organic Contaminants in Air at an Alpine Site Karen S. Lavin,† Kimberly J. Hageman,*,† Samuel K. Marx,‡ Peter W. Dillingham,§ and Balz S. Kamber|| †
Department of Chemistry, University of Otago, Dunedin 9016, New Zealand School of Earth and Environmental Science, University of Wollongong, Wollongong 2522, Australia § George Perkins Marsh Institute, Clark University, Worcester, Massachusetts 01610, United States School of Natural Sciences, Trinity College, Dublin, Dublin 2, Ireland
)
‡
bS Supporting Information ABSTRACT: An approach using trace elements in particulate matter (PM) to identify the geographic sources of atmospherically transported semivolatile organic contaminants (SOCs) was investigated. Daily samples of PM and SOCs were collected with high-volume air samplers from 16 January to 16 February 2009 at Temple Basin, a remote alpine site in New Zealand’s Southern Alps. The most commonly detected pesticides were dieldrin, transchlordane, endosulfan I, and chlorpyrifos. Polycyclic aromatic hydrocarbons and polychlorinated biphenyls were also detected. For each sampling day, the relative contribution of PM from regional New Zealand versus long-range Australian sources was determined using trace element profiles and a binary mixing model. The PM approach indicated that endosulfan I, indeno[1,2,3-c,d]pyrene, and benzo[g,h,i]perylene found at Temple Basin were largely of Australian origin. Local wind observations indicated that the chlorpyrifos found at Temple Basin primarily came from the Canterbury Plains in New Zealand.
’ INTRODUCTION Semivolatile organic contaminants (SOCs) are distributed throughout the globe and have been measured in the Earth’s most remote alpine and polar ecosystems.13 The presence of these toxic, bioaccumulative, and long-lived chemicals in regions where they have never been used or produced is primarily due to transport through the atmosphere.4 Public awareness about the presence of SOCs, including pesticides, polycyclic aromatic hydrocarbons (PAHs), and polychlorinated biphenyls (PCBs), in pristine ecosystems has led to international efforts to reduce their production and use.5 However, further management action requires information about where SOCs in specific remote ecosystems originate. Identifying SOC sources is the first step toward reducing their transport and impact. Until now, local wind direction data from weather stations and/or air mass back-trajectory modeling have been used to gain information about the geographic sources of SOCs at specific local or remote sites.610 Both of these approaches are based on predicting the pathway of airflow to the sampling site. However, neither of these approaches take into account the complex wind systems and mixing processes occurring in remote mountainous areas. Mountains act as barriers to large-scale airflow (terrainforced winds), resulting in airflow being forced around and over r 2011 American Chemical Society
mountains. In many mountainous areas, the airflow that is forced over the mountain descends on the lee side, competing and interacting with meso- to local-scale mountain winds.11 Due to the interaction and mixing of small- and large-scale wind systems, highly variable wind patterns are experienced at alpine sites. Although local wind observations may be useful for describing regional sources, they may fail to adequately describe long-range sources of SOCs due to this high variability. On the other hand, air mass back-trajectory modeling is unable to discern the smallscale diurnal mountain winds, which are responsible for transporting regionally sourced SOCs, and instead focuses on the pathway of the large-scale winds to the sampling site. This is problematic because the contributions of small-scale winds and the mixing of large- and small-scale winds are ignored in source apportionment approaches based on air mass back-trajectories. For these reasons, it is necessary to explore more robust methods for identifying the geographic sources of SOCs in remote alpine ecosystems. Received: August 5, 2011 Accepted: November 9, 2011 Revised: October 20, 2011 Published: November 09, 2011 268
dx.doi.org/10.1021/es2027373 | Environ. Sci. Technol. 2012, 46, 268–276
Environmental Science & Technology
ARTICLE
Figure 1. Map of (a) New Zealand and Australia, (b) New Zealand's South Island showing the Temple Basin sampling site and Canterbury Plains source region, and (c) Australia showing the Australian source region.
The main objective of this study was to investigate a new geographic source apportionment method for SOCs that takes both small- and large-scale wind contributions, and their mixing, into account. The method outlined in this study involves the use of trace element profiles in atmospheric particulate matter (PM) to determine the origin of PM-bound and gaseous-phase SOCs. New Zealand and Australian PM can not only be distinguished from one another on the basis of their trace element composition but also Australian PM sources can be resolved to within 100 km.12,13 Using this method, Marx et al. previously showed that specific regions in Australia act as sources of PM to the West Coast of New Zealand (Figure 1a).12,14 PM and SOCs (both PM-bound and gaseous-phase) are transported through the atmosphere via wind currents15,16 and are removed from the atmosphere by wet and dry deposition.1721 When several wind systems mix, as they do in mountains, the PM and SOCs in each of those systems will also mix. Depending on their specific source regions, SOCs and PM may join the air mass at different stages on route to the sampling site; however, the SOCs and PM in an associated air mass would have been transported via the same wind current and from the same direction. Therefore, we propose that PM can be used as a directional tracer and that the trace element profiles of PM can not only be used to determine PM origin but also to indirectly infer the origin of both PM-bound and gaseous-phase SOCs. The key difference between this SOC source apportionment method and those previously described is that it uses chemical tracers in the air
to understand geographic sources rather than wind measurements or air mass modeling. Additional objectives of this study were to (a) use the PMbased approach to determine which of the SOCs measured at Temple Basin, a remote alpine site in New Zealand’s Southern Alps, were primarily of Australian origin and (b) compare this method with other methods for differentiating sources of SOCs. Daily samples of PM and SOCs were simultaneously collected using two high-volume air samplers operated at Temple Basin from 16 January to 16 February 2009. For each sampling day, the trace element profile of the PM was compared to the trace element profiles of potential Australian and New Zealand PM source regions. From this, the percent Australian PM (% AustPM) was calculated for each day and used in correlation analysis as an indicator of Australian-sourced SOCs. The SOCs targeted for analysis included pesticides (historic-use, currentuse, and recently restricted), PAHs, and PCBs.
’ EXPERIMENTAL SECTION Sampling Site and Air Sampling Method. Air samples were collected at Temple Basin Ski Club (42.91°S, 171.57°E, 1320 m asl), located in Arthur’s Pass National Park in the Southern Alps, New Zealand (Figure 1b). Temple Basin is on a northwesterly facing slope on the west side of the main divide of the Southern Alps. The vegetation is tussock grassland and herbfield, and the site is ∼500 m above tree line (beech forest). This site is ∼60 km from the west coast of the South Island, ∼110 km from the east 269
dx.doi.org/10.1021/es2027373 |Environ. Sci. Technol. 2012, 46, 268–276
Environmental Science & Technology
ARTICLE
coast, and ∼70 km west of the Canterbury Plains, an intensive agricultural region. A weather station (La Crosse Weather Station Pro WS2355, La Crosse Technology, La Crosse WI, USA) was deployed at the site and used to measure wind direction, wind speed, temperature, and barometric pressure. Data were recorded every two minutes. Atmospheric SOCs and PM were collected with two highvolume air samplers, which were operated simultaneously over 31 days, from 16 January 2009 to 16 February 2009. Thus, sampling occurred during the Southern Hemisphere summer when atmospheric SOC concentrations were expected to be at their highest. The sampling media and procedures for sample and field blank collection are described in the Supporting Information. Analytical Method for SOCs. The approach for quantifying SOCs in air sampling media was based on that previously described by Primbs et al.8,22 In brief, air sampling media were individually packed into stainless steel extraction cells and spiked with a mixture of isotopically labeled pesticides, PAHs, and PCBs for use as surrogate standards in the quantification. Analytes were extracted using an Accelerated Solvent Extraction (ASE) System (Dionex, Sunnyvale CA, USA). The extract was concentrated to 0.3 mL and then spiked with three additional isotopically labeled analytes, which were used as internal standards. All sample extracts were analyzed using an Agilent (Santa Clara CA, USA) 6890N gas chromatograph (GC) equipped with an Agilent 5975B mass selective (MS) detector in selective ion monitoring mode. Details of SOCs analysis and instrumental methods are provided in the Supporting Information. Analytical Method for Trace Elements. The analytical method for the determination of trace elements in PM was based on that described by Kamber et al.23 To release PM from polycarbonate filters, filters were folded loosely and placed in 30-mL Teflon beakers, which were filled with triple sub-boiling distilled 5% HNO3. The filters were soaked in this solution at 100 °C for 72 h with intermittent ultrasonic agitation and were then removed with Teflon tweezers and placed back into their respective storage bags. The HNO3 was dried on a hot plate in a clean-room. Next, the filters were lifted from their storage bags and rinsed over their respective open Teflon beakers with ultraclean water. The water was then evaporated and the rinsing repeated. The dried residue was digested with 0.5 mL of a 2:1 mixture of concentrated HF:HNO3. The fluorides in the digest were twice converted to nitrates with 0.25 mL of HNO3 and dissolved in 2% HNO3, which was doped with internal standard containing enriched isotopes (6Li, 103Rh, 187Re, 209Bi, 235U) to correct for matrix suppression and instrumental drift. Prior to analysis, the solutions were heated at 100 °C for 12 h and then centrifuged, after which no visible residues were found. The analyses were performed on an XSeries2 Thermo Fisher Scientific GmbH (Bremen, Germany) quadrupole inductively coupled plasma mass spectrometer (ICP-MS). The ICP-MS protocol was based on that described by Kamber et al.,24 except that an external drift monitor was not used. Two separate digestions of the natural rock USGS reference material, W-2, were used as a calibration standard. Calibration concentrations used for W-2 and the concentrations obtained for the USGS standard, BHVO1 (including long-term averages and the deviation of BHVO-1 concentrations obtained in this study relative to the long-term averages), are provided in Supporting Information, Table S3. Determination of Source Region Indicators. %AustPM. The procedure for determining the proportion of Australian
versus New Zealand PM in a sample has been described in detail previously.12 The main difference in the procedure outlined here is that atmospheric PM was collected on a filter, as opposed to the collection of deposited PM from the surface of a glacier. Twentyone trace elements were selected for the provenance analysis; these elements were assumed to behave conservatively during entrainment, transport, and deposition of the PM. They included rare earth elements (La, Ce, Pr, Nd, Sm, Eu, Gd, Tb, Dy, Y, Ho, Er, Tm, Yb, and Lu), actinoid elements (Th and U), high field strength elements (Nb, Ta and Sc), and a Group 13 metal (Ga). Twenty-five trace elements were used for PM provenancing by Marx et al.;12 however, for this study Co, Cr, Ni, and W were removed because they were found to be affected by anthropogenic activities.13,25 Ti was also removed because the polycarbonate filter blanks contained relatively high Ti. Ga was included in this study because it is a useful element for distinguishing between PM source areas.26 The concentration of the 21 selected elements was expected to reflect mineralogical composition and mass of PM as well as the geochemistry of the particular source areas from where PM was entrained. Provenancing of the PM samples was achieved by determining the major Australian and major New Zealand source region and then using a binary mixing model to determine their relative contributions. First, the profile of 21 trace elements in the sample was normalized and the normalized profile compared to a catalogue of potential source sediments from New Zealand (n = 8) and Australia (n > 200).12,14,27 For each comparison (i.e., each sample compared to each source sediment), the sum of differences (ΣD) was calculated using eq 1
∑ D ¼ ðThPM =Thsource Þ þ ðUPM =Usource Þ þ ::::::::: þ ðGaPM =Gasource Þ
ð1Þ
where ΣD is the sum of differences between each normalized trace element concentration in the sampled PM relative to that in the potential source sediment. The Australian and New Zealand potential source sediments that best matched the sampled PM were identified as those with the lowest ΣD. A binary mixing model was then used to determine the relative contributions of the best-matched Australian and New Zealand PM sources in the sampled PM. The mixing model was used to identify the Australian source contribution at which the ΣD was minimized (Supporting Information, Figure S1). The percent of Australian-derived PM in the total mass of PM in a sample (% AustPM) (Supporting Information, Figure S1) was defined as the relative contribution of the Australian source when ΣD was minimized. A similar approach was successfully used to relate the origin of 210Pb in PM of mixed New Zealand and Australian origin.28 Local Wind Direction Indicators. Wind observations were collected at the sampling site every two minutes. Wind direction observations were reported as one of 16 directions (N, NNE, NE, ENE, S, etc.) and converted to degrees. The frequency of each direction was weighted by multiplying it by wind speed, and these values were used to calculate the percent of time that the wind came from specific source directions. Two potential source regions were identified, one from the northwest (275°355°), representing sources from Australia and the West Coast of New Zealand (%Northwest), and one from the south (125°235°), representing sources from the Canterbury region of New Zealand (%South). Although the wind directions were defined to 270
dx.doi.org/10.1021/es2027373 |Environ. Sci. Technol. 2012, 46, 268–276
Environmental Science & Technology
ARTICLE
correspond to long-range sources in Australia and regional sources in Canterbury, the wind data only gave information about the wind direction experienced at the sampling site. Other regions in New Zealand were not considered as potential sources of SOCs to Temple Basin due to the geography and meteorology of the Southern Alps (Figure 1b). Air Mass Back-Trajectory Indicators. Four-day air mass backtrajectories were generated using the HYSPLIT model (NOAA Air Resources Laboratory, Silver Spring MD, USA). Six backtrajectories were generated for each 24-h period of sampling (at 12 p.m., 4 p.m., 8 p.m., 12 a.m., 4 a.m., 8 a.m.) at each of three different altitudes (2500, 2700, 2900 m asl), giving a total of 18 trajectories per sampling day. The starting altitudes were selected to be higher than Mt. Temple (∼1950 m asl) and nearby Mt Rolleston (∼2270 m asl). A method similar to that described in Primbs et al. was used to calculate back-trajectory source indicators.8 The two major source regions, eastern Australia and the Canterbury Plains, were defined using boxes (Figure 1b and c). The Australian source region box was larger than the Canterbury Plains box due to inaccuracies associated with backtrajectories, which increase with distance from the starting point.29 For each sampling day, the hourly points of the 18 trajectories were plotted. The percent of hourly points that fell within the eastern Australia box was called %Aust, whereas the percent falling in the Canterbury Plains box was called %Cant. Other Meteorological and Environmental Parameters. Hourly precipitation data for the town of Hokitika, which is located west of Temple Basin on the West Coast of New Zealand (Figure 1b), were obtained from the National Climate Database (CliFlo) of New Zealand30 and combined to give daily precipitation for each day. The maximum daily temperatures for several locations in eastern Australia (including Melbourne in the State of Victoria, Macintyre River, Orana, Darling River, Namoi Valley and Sydney in New South Wales, and Darling Downs, Central Highlands, Biloela, and Toowoomba in Queensland) were obtained from the Australian Government Bureau of Meteorology.31 These locations were selected to cover a large geographical range across eastern Australia and to include specific cotton-growing areas. Statistical Analyses. Statistical analysis was conducted with SigmaPlot (Chicago, IL, USA) Version 11.0. Spearman correlation coefficients were determined in all cases to ensure that monotonic correlations were detected and because the Spearman approach is more robust for data sets containing nondetects.32
In New Zealand and Australia, technical endosulfan contains ∼66% endosulfan I and ∼33% endosulfan II isomers.35 Technical endosulfan was banned in New Zealand on 16 January 200936 but was still used in Australia during this study. The use of endosulfan has since been reviewed in Australia and is to be completely banned on 12 October 2012.37 Chlorpyrifos is registered for use in both countries;38,39 however, it is currently under review in Australia.40 Each of the 19 targeted PAHs was detected on at least one day during the sampling period (Supporting Information, Figure S2b). PAHs are produced as a byproduct of combustion so there are many potential sources in both Australia and New Zealand, including bushfires, vehicle exhaust, and domestic wood burners. Each of the 19 targeted PCBs was detected on at least one of the sampling days (Supporting Information, Figure S2c). PCBs are classified as persistent organic pollutants under the Stockholm Convention.5 The phase-out of PCBs began in New Zealand in 1980s and in Australia in the 1970s; however, equipment containing PCBs is still in use in both countries.34,41 The atmospheric concentrations of dieldrin, the sum of all technical chlordane isomers (Σchlordane), endosulfan I, and chlorpyrifos as well as the sum of all detected PAH isomers (total PAHs) and all detected PCB congeners (total PCBs) are shown for each day of the study in Figure 2. Dieldrin had the highest pesticide concentrations in air (103 pg m3), which is likely due to the historic use of dieldrin in Australasia and the presence of dieldrin in the soils of contaminated sheep-dip sites in New Zealand.42,43 The maximum Σchlordane, endosulfan I, and chlorpyrifos concentrations were 18 pg m3, 12 pg m3, and 12 pg m3, respectively. Of the PAHs, phenanthrene had the highest maximum concentration (1285 pg m3), followed by pyrene (968 pg m3), acenaphthene (662 pg m3), acenaphthylene (645 pg m3), and fluorene (599 pg m3). PCB-101 and PCB-118 had the highest concentrations of the PCB congeners (12 pg m3 and 13 pg m3, respectively). The SOC concentrations measured in this study were comparable to those reported in other mountain locations, with the exception of those reported for dieldrin and endosulfan I. Dieldrin concentrations were high compared to those reported in the Oregon Cascades (
