Environ Monit Assess (2008) 144:351–366 DOI 10.1007/s10661-007-9998-2
Characterization of fine aerosol and its inorganic components at two rural locations in New York State Ramya Sunder Raman & Philip K. Hopke & Thomas M. Holsen
Received: 10 March 2007 / Accepted: 19 September 2007 / Published online: 10 November 2007 # Springer Science + Business Media B.V. 2007
Abstract Samples of PM2.5 were collected to measure the concentrations of its chemical constituents at two rural locations, Potsdam and Stockton, NY from November 2002 to August 2005. These samples were collected on multiple filters at both sites, every third day for a 24-h interval with a speciation network sampler. The Teflo® filters were analyzed for PM2.5 mass by gravimetry, and elemental composition by Xray fluorescence (XRF). Nylasorb® filters and Teflo® filters were leached with water and analyzed for anions and cations, respectively, by ion chromatography (IC). Fine particulate matter (PM2.5) mass and its
Electronic supplementary material The online version of this article (doi:10.1007/s10661-007-9998-2) contains supplementary material, which is available to authorized users. R. Sunder Raman : P. K. Hopke (*) Department of Chemical and Biomolecular Engineering, Clarkson University, Potsdam, NY 13699-5708, USA e-mail:
[email protected] T. M. Holsen Department of Civil and Environmental Engineering, Clarkson University, Potsdam, NY 13699-5708, USA R. Sunder Raman : P. K. Hopke : T. M. Holsen Center for Air Resources Engineering and Science, Clarkson University, Potsdam, NY 13699-5708, USA
inorganic component measurements were statistically characterized, and the temporal behavior of these species were assessed. Over the entire study period, PM2.5 mass concentrations were lower at Potsdam (8.35 μg/m3) than at Stockton (10.24 μg/m3). At both locations, organic matter (OM) was the highest contributor to mass. Sulfate was the second highest contributor to mass at 27.0% at Potsdam, and 28.7% at Stockton. Nitrate contributions to mass of 8.9 and 9.5% at Potsdam and Stockton, respectively, were the third highest. At both locations, fine PM mass exhibited an annual cycle with a pronounced summer peak and indications of another peak during the winter, consistent with an overall increase in the rate of secondary aerosol formation during the summer, and increased partitioning of ammonium nitrate to the particle phase and condensation of other semivolatiles during the winter, respectively. An ionbalance analysis indicated that at both locations, during the summers as well as in the winters, the aerosol was acidic. Lognormal frequency distribution fits to the measured mass concentrations on a seasonal basis indicated the overall increase in particle phase secondary aerosol (sulfate and SOA) concentrations during the summers compared to the winters at both locations. Keywords PM2.5 . Potsdam . Stockton . Mass . Crustal elements . Ions . Lognormal distribution
352
Introduction Subsequent to the promulgation of the 1997 National Ambient Air Quality Standards (NAAQS), intense research activity has focused on determining the sources of PM2.5, understanding its atmospheric dynamics, and transport, with a view of better addressing the challenging problem of fine PM management. The current NAAQS for ground level PM2.5 are set at 15 μg/m3 (3 year average, weighted annual mean) and 35 μg/m3 (24 h, 3 year average, 98th percentile). However, unlike other criteria pollutants like ozone or carbon monoxide, fine particles are a mixture of many compounds that come from a variety of sources. Ambient particles consist of nonvolatile, and semi-volatile species. Additionally, fine particles are predominantly hygroscopic and the mass fraction of water in the condensed phase increases with relative humidity. Typically, at relative humidities of around 80%, water accounts for almost half the fine PM mass (McMurry 2000). Therefore, fine PM composition is intimately linked with the gas phase composition. Thus, quantifying the fine particle mass concentration in itself is quite challenging, and having achieved that, the mass alone is insufficient to provide an understanding of the chemical nature and origin of these particles. Hence, a detailed understanding of the composition of fine particulate matter is necessary in order to identify the emission sources and their relative importance to plan control strategies, and to achieve the objective of regulatory compliance.
Environ Monit Assess (2008) 144:351–366
In this study, samples were collected at two rural locations in New York State, Potsdam and Stockton to determine the concentrations of PM2.5 mass, its inorganic, and carbonaceous components. Often, an analysis of the data reveals the underlying relationships among the various components of fine particulate matter, and provides insights about their sources and formation mechanisms. This paper characterizes fine PM and its inorganic components collected at the study locations between November 2002 and August 2005. A detailed investigation of the carbonaceous fraction of fine particles collected at the study locations during the same period is presented by Sunder Raman and Hopke (2007).
Sampling and chemical analyses Sampling Sampling was carried out at two rural locations in New York state, Potsdam and Stockton (Fig. 1). Samples were collected every third day at both locations by exposing multiple 47 mm filters in an Andersen RAAS2.5-400 speciation network sampler. The Potsdam site (Latitude 44° 40′ 20″ N, Longitude 74° 59′ 16″ W) is located in St. Lawrence County in the northern region of NY near the Canadian border. This site was chosen because it is not significantly impacted by local sources (Hopke et al. 2003). The Stockton site (Latitude 42° 17′ 59″ N, Longitude 79° 23′ 42″ W) is situated in Chautauqua County, approximately 20 km south of Fredonia, NY and approximately 10 km from the eastern shore of Lake Erie. This site is located in a remote rural area, on top of a ridge with an elevation of 488 m. This site was chosen because it is situated away from metropolitan areas. Further, it was well situated to sample the prevailing southwesterly winds from the Midwestern region of the United States entering New York State (Hopke et al. 2003; Sunder Raman and Hopke 2006, 2007). Chemical analyses
Fig. 1 Map of the sampling locations (Produced using Google Earth™)
Analytical techniques were chosen before the commencement of sampling. The methods chosen were based on their scope and applicability, and availability of instrumentation. Samples from Channel 4 of the RAAS sampler were used for mass determination.
Environ Monit Assess (2008) 144:351–366
PM2.5 mass was determined by weighing the filters before and after the sampling in accordance with the requirements specified in 40 CFR 50, Appendix K. Mass measurements were made on the Sartorius MC5 micro analytical balance. Elemental analysis was performed on the Spectro® X-Lab Pro. The instrument was equipped with Barkla scatter, and secondary targets. Samples from Channel 1 of the RAAS were analyzed to determine the concentrations of trace elements using X-ray fluorescence (XRF) analysis. In this study, elements from Na to Cl were analyzed on the Barkla scatter highly oriented pyrolitic graphite (HOPG) target, while Mo to Ce were analyzed on the Al2O3 Barkla scatter target. Barkla scatter targets provide polarized X-rays to reduce the background caused by scattering of the excitation radiation on the sample in the direction of the detector (Spectro® X-Lab manual, Version 04/ 2000). Elements from K to Mn were analyzed on the Co secondary target, while Mn to Zr were analyzed on the Mo secondary target. Secondary targets provide an opportunity to choose the most appropriate excitation condition for the target analytes. Nylasorb and Teflo® filters were leached with water and analyzed for anions and cations, respectively by IC. The Nylasorb® filters from Channel 2 of the RAAS were extracted with 30 ml ultra pure water. Analysis was performed on a Dionex DX-500 ion chromatography system. Isocratic separation was performed using an Ionpac AS4A-SC column and 2.7 mM Na2CO3/0.3 mM NaHCO3 eluent. The details of the procedure are described in Research Triangle Institute (2003a) ‘Standard Operating Procedure for PM2.5 Anion Analysis.’ After the completion of mass, and BC determination, Teflo® filters, were used for cation analysis. The samples were extracted using 30 ml ultra pure water and 100 μl pure ethanol. Analysis was performed on a Dionex DX-500 ion chromatography system. Gradient separation was performed using an Ionpac CS12A column and 22 mM H2SO4 eluent. The details of the procedure are described in the Research Triangle Institute (2003b) ‘Standard Operating Procedure for PM2.5 Cation Analysis.’ Lab and field blanks were collected for mass, element, and ion samples. During the entire sampling period, a lab blank was analyzed every month, and field blanks were collected and analyzed every other week. The blank concentrations were below detection
353
limits for mass and all of the ions. The blank concentrations were significant for all of the XRF elements. Thus, all the elemental concentrations were blank corrected. In the following sections, elemental concentrations correspond to blank corrected corrections. The quantification details are discussed by Sunder Raman (2006).
Results and discussion Descriptive statistics The concentrations of various species, and key statistics for each of these species measured at Potsdam and Stockton are summarized in Tables 1 and 2, respectively. These statistics provide an indication of the relative importance of each species in terms of its contribution to PM2.5. In these tables, BDL indicates the number of values below detection limit. In these tables, only species with a signal to noise ratio (S/N) greater that 0.2 are shown. S/N was determined in accordance with the procedure described by Paatero and Hopke (2003). Consider a concentration data matrix X of the order m × n. Let xij represent an element of the matrix X. Paatero and Hopke (2003) defined S/N ratio for censored data based on the number mDLj of below detection limit values in column j. If the (average) detection limit for censored values in column j is denoted by δj, then, S/ P N ratio is defined as fijxij > δjgxij=δjmDLj. A S/N ratio less that 0.2 indicates a “bad” variable. At Potsdam, among the inorganic species, sulfate was the most abundant species by mass, followed by nitrate, and ammonium ions (Table 1). Among the crustal elements, Si was the most abundant by mass followed by Ca. Cu, Fe, and Zn were important contributors to the transition metal concentrations with Cu having the highest concentration followed by Fe and Zn. The median concentrations of all species except Na, Mg, K, Ti, V, and Mo were higher at Stockton than at Potsdam (Table 2). At Stockton too, sulfate was the most abundant species followed by nitrate and ammonium ions. Si concentration was the highest among crustal elements, followed by Ca. Fe, Cu, and Zn were the highest contributors to the transition metals concentration. At Stockton median Fe concentration was higher than the median concentrations of Cu and Zn.
354
Environ Monit Assess (2008) 144:351–366
Table 1 Summary statistics of 30 species and PM2.5 mass concentration at Potsdam Concentration (ng/m3)
N=345
Species
Mean
Median
Geometric mean
Maximum
Minimum
% BDL
% Missing
S/N
Na Mg Al Si P S Cl K Ca Ti V Cr Fe Co Ni Cu Zn Se Sr Mo Cd Sn Sb W Pb SO2 4 NO 3
95.8 20.1 32.3 81.2 5.5 658.2 33.9 47.1 37.8 9.6 2.3 9.1 35.4 5.3 16.1 39.6 27.6 2.7 13.7 37.0 11.6 56.9 53.7 17.7 19.2 2,251.4 736.2 107.0 492.1 43.3 8,350.0
61.1 14.8 26.8 66.5 4.6 405.8 18.0 45.0 30.3 6.8 1.6 6.5 28.0 3.6 13.6 31.5 23.8 2.1 12.1 33.2 9.6 50.0 61.2 16.5 17.7 1,291.9 399.4 82.3 286.7 35.5 5,966.0
63.8 15.4 26.0 68.6 4.8 427.3 23.0 38.6 31.6 6.6 2.0 6.8 28.5 4.2 14.2 30.6 24.5 2.2 13.0 30.6 10.2 50.2 43.7 17.0 18.4 1,449.2 397.0 78.0 236.4 35.9 6,198.0
1,003.6 154.0 297.3 833.7 24.4 6,285.1 479.3 176.7 367.5 153.4 6.0 404.5 333.0 95.1 392.5 657.7 500.7 25.4 49.6 654.1 43.7 1,439.0 87.6 435.3 409.1 19,112.9 9,278.1 521.5 4,175.3 319.2 49,267.0
10.9 3.6 10.4 19.7 3.2 14.1 10.2 13.3 8.5 0.8 1.8 5.4 10.3 2.5 9.7 8.8 4.1 0.9 10.1 6.2 5.0 13.1 4.3 10.9 12.0 41.6 26.1 21.8 22.4 18.1 425.0
5.5 34.1 22.0 17.6 52.9 0.4 20.2 19.7 4.3 4.6 69.4 19.1 20.5 42.8 19.1 6.1 9.5 60.4 25.2 33.2 43.1 13.9 76.6 19.4 19.1 1.9 1.9 3.5 5.5 19.1 0.0
14.2 14.2 14.2 14.2 14.2 14.2 14.2 14.2 14.2 14.2 14.2 14.2 14.2 14.2 14.2 14.2 14.2 14.2 14.2 14.2 14.2 14.2 14.2 14.2 14.2 13.1 13.1 17.1 17.1 17.1 15.7
127.9 9.5 13.6 31.4 0.7 10,339 5.5 13.9 84.8 228.4 0.3 6.0 11.1 0.5 24.3 59.8 78.2 1.1 1.7 21.4 1.2 387.0 0.7 34.5 3.6 4,319.0 7,578.0 447.9 2,907.0 2.6 –
Na+ NHþ 4 K+ PM2.5
Temporal behavior of PM2.5 and its inorganic components For this study, fine PM inorganic components were grouped into three broad categories. They included crustal elements (Si, Al, Ca, Fe and Ti), transition metals (V, Cr, Ni, Cu, Zn) and Pb, and inorganic ions + + þ (SO2 4 , NO3 , Na , NH4 , and K ). PM2.5 mass The monthly average PM2.5 concentration and ambient temperature time series at both locations are shown in Fig. 2. The error bars represent one standard deviation in the measured values. The mass concen-
tration time series at both locations exhibit a seasonal behavior with two annual peaks (Fig. 2). One peak occurs during the summer (May to August), while there are indications of another peak during the winter (November to the following February). However, during 2002 winter at both locations two peaks were observed, one in December 2002 and the other in March 2003. It can also been seen that the monthly average temperatures during winter 2002/03 were lower than the winter temperatures during 2003/04, and 2004/05 (Fig. 2). An explanation for the observation of two peaks during winter of 2002, instead of one is currently unknown. It must also be noted that the winter peaks were less pronounced compared to the summer peaks at both locations.
Environ Monit Assess (2008) 144:351–366
355
Table 2 Summary statistics of 30 species and PM2.5 mass concentration at Stockton Concentration (ng/m3)
N=344
Species
Mean
Median
Geometric mean
Maximum
Minimum
% BDL
% Missing
S/N
Na Mg Al Si P S Cl K Ca Ti V Cr Fe Co Ni Cu Zn Se Sr Mo Cd Sn Sb W Pb SO2 4 NO 3
67.1 16.0 34.1 83.6 5.7 1,143.1 22.1 41.9 34.0 4.0 1.5 6.8 43.9 4.7 14.6 33.6 29.7 3.8 13.9 34.1 12.7 52.9 59.5 17.9 20.6 2,968.8 974.0 120.6 673.4 52.2 10,236
58.8 14.0 28.2 78.9 4.6 836.6 18.3 35.0 31.4 3.6 1.3 6.6 37.6 3.6 13.9 32.0 26.5 2.7 12.5 33.0 9.9 52.5 65.1 17.2 18.5 1,969.9 577.8 124.7 450.3 37.9 7,858.0
53.9 14.1 26.3 75.7 5.0 748.4 20.0 32.7 31.6 3.6 1.5 6.7 37.4 4.0 14.3 32.9 27.6 2.9 13.2 32.9 10.9 51.8 54.2 17.6 19.5 1,625.8 562.0 95.7 383.4 39.8 7,557.0
246.8 40.3 400.0 211.8 24.6 6,522.4 148.9 339.5 87.4 12.5 2.5 24.9 165.2 24.3 29.8 67.7 101.9 19.9 50.3 64.9 50.8 98.9 81.7 49.4 78.4 20,144.1 7,023.8 274.7 3,979.4 822.9 44,169.0
10.2 4.6 6.2 25.2 3.1 11.3 10.4 11.5 9.6 0.8 1.1 4.3 11.3 2.5 10.3 14.2 4.3 1.1 10.2 11.3 5.1 28.8 7.0 12.8 11.6 23.0 37.7 21.0 16.3 18.9 439.0
5.2 13.4 5.1 3.1 29.9 0.4 3.6 8.0 2.5 2.8 39.1 3.1 5.1 16.5 3.1 2.8 4.2 21.7 3.9 14.5 23.2 11.2 43.5 3.1 3.1 1.6 1.0 6.0 1.1 11.3 0.0
50.0 50.0 50.0 50.0 50.0 50.0 50.0 50.0 50.0 50.0 50.0 50.0 50.0 50.0 50.0 50.0 50.0 50.0 50.0 50.0 50.0 50.0 50.0 50.0 50.0 45.4 45.4 52.9 52.9 52.9 50.6
97.1 13.4 43.4 128.2 0.8 12,955 14.1 19.2 77.9 93.2 0.2 19.5 38.0 1.0 96.6 65.9 114.2 4.7 8.4 31.2 1.5 240.0 0.5 156.7 16.9 1,959.7 1,866.1 12.0 610.7 2.9 –
Na+ NHþ 4 K+ PM2.5
A comparison of the patterns exhibited by the PM2.5 concentration and temperature time series at both locations reveal the relationships between temperature driven rates of formation of secondary species, and the extent of gas-to-particle partitioning. The overall rate of secondary aerosol (both inorganic and organic) formation increases during the summer compared to the winter. However, during the summers most of ammonium nitrate is partitioned on to the gas phase unlike during the winters. Thus, it is hypothesized that the summer peak in mass concentration corresponds to the increase in secondary sulfate, and secondary organic aerosol (SOA) concentrations. During the winters, the lower temperatures promote increased partitioning of ammonium nitrate to the
particle phase, accounting for the observed winter peak in mass concentrations during the cooler months. Additionally winter fine PM concentration may also increase due to wood and oil burning for residential heating, and condensation of semi-volatile organic species. In order to examine the relationship between temperature and mass concentration, these variables were regressed at both locations. The overall r2 at Potsdam was 0.17. However, when only the summer months were considered, the r2 value was 0.52. This value of r2 indicates that mass concentration was moderately correlated with temperature during the summers. Likewise, at Stockton the overall r2 and summer r2 were 0.37 and 0.69, respectively. No
356
Environ Monit Assess (2008) 144:351–366
Fig. 2 Monthly average PM2.5 concentration and temperature time series Potsdam (top), and Stockton (bottom)
significant correlation was observed when only the winter months were considered for the regression analysis (r2 =0.01, and 0.02 at Potsdam and Stockton, respectively). These observations suggest that secondary formation of fine PM during summers is an important contributor to the overall PM2.5 burden at both locations. In order to assess the impact of activities that vary as a function of the day-of-week on measured mass concentrations at both locations, a day-of-week analysis was performed. Figure 3 summarizes the day-of-week variations in mass at both locations. The boxes contain data from the 25th to the 75th percentile. The 10th and 90th percentile limits (whiskers) and the 5th and 95th percentile outliers, as well as the median (solid line), and mean (dotted line) are also shown in the plot. At Potsdam, little variation was observed in the median concentrations from day-to-day, except for a slight decrease on Sunday. At Stockton, there was more variability in the daily median concentrations. How-
ever, no marked decrease was observed in the median concentrations during the weekend (Saturday and Sunday) compared to the weekdays at both locations. These observations suggest that local source impacts on PM concentrations at both locations are minimal. Both sites are located away from major cities, and are not influenced by urban traffic patterns that often exert significant influence on the weekday/weekend PM concentration trends. A regression analysis of PM2.5 mass concentrations measured at both locations revealed that the overall inter-site coefficient of determination (r2) was 0.54, while the summer r2 was 0.59, and winter r2 was 0.12. The r2 values indicate that the air flow regime during the summers provides similarly processed air to both locations more often during the summers compared to the winters. An alternative interpretation for the r2 values is that PM2.5 concentrations at both locations are dominated by regional and local sources during the summer and winter, respectively.
Environ Monit Assess (2008) 144:351–366
357
Fig. 3 Day-of-week variations in PM2.5 concentration at Potsdam, and Stockton
Crustal elements The concentration time series of crustal elements as monthly averages at both locations are shown in Fig. 4. The error bars represent one standard deviation in the measured values. At Potsdam, with the exception of Ti (a concentration peak only during 12/2004), all other the elements exhibit an annual maximum between March and May, and a minimum between October and December (Fig. 4 (top)). At Stockton too, a similar pattern was exhibited (Fig. 4 (bottom)). A clear increasing trend between March and May can be observed. However, due to sampling gaps, the minimum for 2003 is less obvious. The maximum concentration observed between March and May, may be due to favorable meteorological conditions during spring that aid the transport of fine airborne soil. Transported Asian dust (Liu et al. 2003a) events may also partly contribute to the increase in soil concentrations during spring. Five day back trajectories calculated using the NOAA HYSPLIT model (Draxler and Rolph 2003) on several high crustal element concentration days (e.g., 03/25/2004, 04/30/2004, 03/ 26/2005) at both locations indicated the likely transport of Asian dust to Potsdam and Stockton. Further, plowing of agricultural fields during the spring season also results in increased resuspension of soil. The minimum concentration observed between October and December may be due to less arid
conditions preventing the efficient resuspension and transport of airborne soil. Table 3 summarizes the coefficients of determination values between crustal elements at Potsdam and Stockton. All the r2 values are significant at the 95% confidence level. At Potsdam, Al was reasonably well correlated with Si, and Ca (Table 3). Si was also well correlated with Ca. Correlations between Si and Fe, and Ca and Fe were moderately high. All other correlations between crustal elements were moderate to poor. The strength of the relationship between Si and Ca was the strongest, while that between Al and Ti was the weakest. At Stockton, the correlations between all elements were weaker than those at Potsdam, except for those between Ti and Al, and Ti and Fe (Table 3). At Stockton too, the strength of the relationship between Si and Ca was the strongest. The inter-site correlations are summarized in Table 4. In this table, P and S in the parentheses indicate Potsdam and Stockton, respectively, and ‘x’ indicates that r2 was not significant at the 95% confidence level. Every element at Potsdam was either uncorrelated or weakly correlated with its corresponding pair at Stockton, except for moderate correlations between Al and Si. Thus, the r2 values indicate that the air flow regimes provide different air masses to the two sites. An additional interpretation of the r2 values is that on average, the soil contributions at each of these locations are likely to be dominated by local sources.
358
Environ Monit Assess (2008) 144:351–366
Fig. 4 Monthly average concentration time series of Al, Si, Ca, Ti, and Fe at Potsdam, and Stockton
Environ Monit Assess (2008) 144:351–366
359
Table 3 Coefficient of determination between crustal element concentrations at Potsdam and Stockton Rural locations Potsdam N=226 Al Si Ca Ti Fe Stockton N=155 Al Si Ca Ti Fe
Al 1
Si 0.76 1
Ca 0.69 0.79 1
Ti 0.32 0.37 0.48 1
Fe 0.52 0.64 0.63 0.35 1
Al 1
Si 0.56 1
Ca 0.35 0.62 1
Ti 0.44 0.55 0.48 1
Fe 0.37 0.48 0.48 0.53 1
Transition metals and Pb The concentration time series of transition metals and Pb as monthly averages at both locations are shown in Fig. 5. The error bars represent one standard deviation in the measured values. No marked seasonal behavior is observed for the transition metals and Pb concentration at Potsdam (Fig. 5 (top)). However, two peaks in V concentrations were observed during 02/2003 and 12/2004. Further, there are several other V peaks of lower magnitude observed throughout the sampling interval. Thus, the V peaks appear to be random. A concentration peak for all the transition metals was observed during 07/2004 at Potsdam (Fig. 5 (top)). On 07/23/2004 all transition metals (except V) and Pb exhibited unusually high concentrations. The 10 day back trajectory beginning on 07/23/2004 calculated using the NOAA HYSPLIT model (Draxler
and Rolph 2003) indicated that the air mass traveled across Ontario, Canada, and the Ohio River Valley. Thus, the back trajectories were uninformative from the perspective of identifying a specific source of transition metals. However, it is suggested that transition metals may have accumulated during the passage of the air parcel across the industrialized regions of Canada, and the US before it reached Potsdam. Thus, 07/2004 for transition metals peak is directly linked to the high concentration event on 07/ 23/2004. At Stockton too, several random peaks in V concentration are observed (Fig. 5 (bottom)). Again, no marked seasonal behavior is observed for the transition metal concentrations. At Potsdam moderate correlations between Cr and Ni, and Ni and Pb are observed (Table 5). This observation indicates the likelihood of a common source for Ni and Cr, and Ni and Pb, respectively. Liu et al. (2003b) had resolved a Ni smelter source at Potsdam. All metals at Stockton were weakly correlated except for Ni and Pb which were moderately correlated (Table 5). However, Liu et al. (2003b) did not resolve a Ni smelter source at Stockton. The inter-site correlations between transition metals, and Pb at Potsdam and Stockton were also examined. Zn concentrations at Potsdam were weakly correlated with Cu concentrations at Stockton with a r2 value of 0.21. For all other combinations, the coefficient of determination values were not significant at the 95% confidence level. Liu et al. (2003b) resolved Zn and Cu smelters at both locations. However, the poor correlations between the Cu and Zn concentrations at the two locations can be explained by the fact that the PSCF analysis by Liu et al. (2003b) indicated influence from different specific sources at these locations. Inorganic ions
Table 4 Inter-site (Potsdam and Stockton) coefficient of determination between crustal elements N=113
Al (S)
Si (S)
Ca (S)
Ti (S)
Fe (S)
Al (P) Si (P) Ca (P) Ti (P) Fe (P)
0.41 0.36 0.11 0.01 0.16
0.22 0.42 0.20 0.01 0.21
0.05 0.19 0.13 x 0.09
0.16 0.25 0.10 x 0.12
0.10 0.22 0.10 x 0.15
The monthly average concentration time series of the + þ anions (SO2 4 and NO3 ), and cations (Na , NH4 and + K ) throughout the sampling period at Potsdam and Stockton are shown in Fig. 6. The error bars represent one standard deviation in the measured concentrations. Characteristic annual summer peaks for sulfate, and winter peaks for nitrate are observed at both locations (Fig. 6). Further, an examination of sampleto-sample variations (Supplemental Fig. 1) indicated that nitrate and ammonium concentrations showed
360
Environ Monit Assess (2008) 144:351–366
Fig. 5 Monthly average concentration time series of V, Ni, Cr, Cu, Zn and Pb at Potsdam, and Stockton
several co-incident peaks between November and February (winter) every year, although there were year-to-year variations. Sodium ion showed concentration peaks throughout the sampling period. Several
peaks were observed during the winters at both locations (Fig. 6). However, overall the time series appeared to be random and there were no marked temporal patterns.
Environ Monit Assess (2008) 144:351–366
361
Table 5 Coefficient of determination between transition metals, and Pb concentrations at Potsdam and Stockton Rural locations Potsdam N=191 V Cr Ni Cu Zn Pb Stockton N=144 V Cr Ni Cu Zn Pb
V 1
Cr 0.18 1
Ni 0.05 0.58 1
Cu 0.04 0.08 0.09 1
Zn 0.09 0.23 0.27 0.22 1
Pb 0.06 0.37 0.67 0.05 0.31 1
V 1
Cr 0.07 1
Ni 0.04 0.05 1
Cu 0.02 0.06 0.14 1
Zn 0.18 0.16 0.18 0.10 1
Pb 0.04 0.08 0.67 0.13 0.20 1
At both locations the highest monthly average potassium ion concentration was observed in August 2003 (Fig. 6). In August 2003, as many as 900 wildfires occurred in Canada (Natural Resources Canada, Canadian Forest Services, 2006). Canadian Forest Service, Natural Resources, Canada reported “out of control fires” in Ontario and Manitoba provinces of Canada on various occasions during the month of August 2003. Thus, the transport of fine aerosol from these fires may have been responsible for the elevated potassium concentrations. Further, water soluble potassium concentrations are also linked to wood combustion. Wood combustion during winters for residential heating can explain the concentration peaks observed during the winters (November to February) at both locations (Fig. 5). At Potsdam, a weak correlation was observed þ between NO 3 and NH4 (Table 6). This observation can be explained by the fact that during winters the lower ambient temperatures promote ammonium nitrate formation. The correlation between all the other ions was poor. At Stockton the correlation between NHþ 4 and NO3 was weaker than that at + Potsdam, while that between NHþ was 4 and K slightly stronger (Table 6).
The inter-site correlations between inorganic ions are summarized in Table 7. In this table, x indicates that r2 was not significant at the 95% confidence level. P and S in parentheses indicate Potsdam and Stockton, respectively. Inter-site correlations between SO2 concentrations, Na+ concen4 þ trations, and NH4 concentrations are moderate, + while that between NO 3 concentrations and K þ concentrations are weak. Although weak, NH4 concentrations at one location and NO 3 concentrations at the other are also correlated. The correlation between inter-site SO2 concentrations 4 is the strongest. This observation may be due to the regionally transported sulfate being common to both locations. Seasonal variations in ion balance The seasonal variations in the acidity of fine PM at both locations were examined. The effect of minor inorganic ions and organic ions were neglected. However, the use of major inorganic anion concentrations for ion balance is expected to yield a reasonable measure of the actual fine PM acidity (Schwab et al., 2004). An ion balance calculation using the concentrations of major inorganic ions þ (SO2 4 , NO3 , and NH4 ) was performed. The results were normalized by mass, expressing ion balance in units of milli equivalents per gram (m equiv/g). For the seasonal analysis, winter was defined as the period from the beginning of November through the end of the following February, and summer as the period from beginning of May through the end of August. The seasonal variations in ion balance at both locations are summarized in Fig. 7. In this figure, a negative value indicates that the aerosol is acidic, zero indicates neutral aerosol, and a positive value indicates alkaline aerosol. As the ion balance becomes more and more negative, the aerosol acidity increases. The range of average ion balance values at Potsdam and Stockton agree well with the results reported by Schwab et al. (2004) for the Whiteface Mountains, NY in a study conducted between 2001 and 2003. Schwab et al. (2004) reported ion balance values ranging from approximately −1.3 to − 4.2 m equiv/g at the Whiteface Mountains for different seasons during their study period.
362
Environ Monit Assess (2008) 144:351–366
Fig. 6 Monthly average concentration time series of , , Na+, , and K+ at Potsdam, and Stockton
At Stockton, the acidity during the winters is lower than that during the summers, except for the winter 2003 and summer 2004 acidities which were comparable (Fig. 7). The observations made at
Stockton must be interpreted carefully because few observations were available during winter 2004, and summer 2005. At Potsdam, although the summer acidity every year was higher than its preceding
Environ Monit Assess (2008) 144:351–366
363
Table 6 Coefficient of determination between inorganic ion concentrations at Potsdam and Stockton
Fitting distribution functions to PM2.5 mass
Rural locations
Many probability distribution functions (pdfs) such as the lognormal distribution, log-logistic distribution, gamma distribution, and Weibull distribution have been used to fit air pollutants concentration data (Seinfeld and Pandis 1998; Lu 2002; Karaca et al. 2005). The choice of the pdf depends on the purpose of fitting a pdf to mass concentrations. Usually the best fit pdf is chosen when it is desired to predict future concentrations, or exceedences. A lognormal distribution was fit to the mass measurements in order to understand and assess the importance of the chemical processes governing the formation of fine PM. Lognormal distributions have been successfully fitted to PM10, fine PM chemical components, and gas-phase pollutants in Duarte, CA. The pdf was found to be approximately unimodal lognormal (Kao and Friedlander 1994, 1995). A study of the crustal elements in Rubidoux, CA revealed that Al and Si frequency distributions were bimodal lognormal distribution. The lognormal distribution pattern exhibited by air pollutants has been attributed to the theory of random successive dilutions (Ott 1990). Further, it was also suggested that deviations in the measured air pollutant concentration from the lognormal distribution may be due to sampling errors (deNevers et al. 1979). Kao and Friedlander (1994, 1995) suggested that the width of geometric standard deviation of the fitted distributions increased with increasing reactivity of the aerosol component. In this study, the same concept was extended to examine the variations in aerosol reactivity between seasons. The two-parameter lognormal distribution function is defined as follows:
Potsdam N=223 SO2 4 NO 3 Na+ NHþ 4 K+ Stockton N=119 SO2 4 NO 3 Na+ NHþ 4 K+
SO2 4 1
NO 3 x 1
Na+ x x 1
NHþ 4 0.19 0.34 X 1
K+ 0.10 0.08 0.03 0.22 1
SO2 4 1
NO 3 x 1
Na+ 0.07 x 1
NHþ 4 0.19 0.17 X 1
K+ 0.08 0.01 0.08 0.27 1
winter, the summers high, winter low pattern in acidity is not well established (Fig. 6). For example, although summer 2003 acidity was higher than winter 2002, it was lower than the winter 2003 acidity. However, at Potsdam a steadily increasing trend in the aerosol acidity between winter 2002 and summer 2005 was observed. During the seasons when sufficient data points were available at Stockton to make reliable conclusions, the acidity at Stockton during each season was mostly higher than the corresponding acidity in Potsdam. This observation is indicative of the fact that Potsdam may have more sources of ammonia compared to Stockton due to a large number of diary farms being located in the vicinity of Potsdam. Increased availability of ammonia increases the atmospheric neutralizing ability, and results in a less acidic aerosol at Potsdam, than at Stockton.
Table 7 Inter-site (Potsdam and Stockton) coefficient of determination between inorganic ions N=83
þ + + SO2 4 ðSÞ NO3 ðSÞ Na (S) NH4 ðSÞ K (S)
SO2 4 (P) NO 3 (P) + Na (P) NHþ 4 (P) K+ (P)
0.48 x x 0.12 0.06
x 0.24 x 0.26 x
0.08 x 0.49 x x
0.19 0.21 x 0.37 0.08
0.13 x x 0.10 0.29
1 ðln x ln mÞ pð xÞ ¼ pffiffiffiffiffi exp 2 ln2 s 2p x ln s where μ and σ are the geometric mean, and geometric standard deviation respectively. The goodness of the fit was tested by the Kolmogorov Smirnov test (K-S) test (Kao and Friedlander 1994, 1995). The K-S test statistic compares the cumulative distribution function of the measurements with that obtained from the theoretical distribution function. The results of this test are characterized by the value of D, which is defined as the maximum of the
364
Environ Monit Assess (2008) 144:351–366
Fig. 7 Seasonal variations in the ion balance at Potsdam, and Stockton
Table 8 Lognormal distribution parameters for PM2.5 mass concentrations at Potsdam, and Stockton
Rural areas Potsdam N 108 94 47 35 37 25 38 23 34 Stockton 57 68 22 25 29 22 26 12 14
Period 01/03–12/03 01/04–12/04 01/05–08/05 Winter 2002 Summer 2003 Winter 2003 Summer 2004 Winter 2004 Summer 2005
pdf parameters μ (ng/m3) 6,570 5,472 7,194 6,210 8,558 4,766 6,828 5,527 7,700
σ 2.04 2.13 2.05 1.78 2.10 2.03 2.28 1.81 2.07
K-S test statistic D p 0.05 0.95 0.08 0.52 0.06 0.99 0.09 0.82 0.09 0.88 0.11 0.91 0.07 0.98 0.10 0.99 0.09 0.93
01/03–12/03 01/04–12/04 01/05–08/05 Winter 2002 Summer 2003 Winter 2003 Summer 2004 Winter 2004 Summer 2005
7,992 7,061 8,643 5,957 9,890 5,193 9,600 4,950 12,092
2.07 2.12 2.28 2.08 2.26 1.82 1.90 1.89 2.00
0.05 0.07 0.09 0.16 0.11 0.20 0.10 0.14 0.20
0.96 0.87 0.99 0.56 0.82 0.34 0.96 0.98 0.64
Environ Monit Assess (2008) 144:351–366
absolute difference between the actual and the theoretical cumulative distribution functions. D is a measure of the disproof of the hypothesis that the distribution is lognormal. The estimated significance level of the K-S statistic, D is indicated by p. The lognormal distribution function was fitted to the measured fine PM mass concentrations at Potsdam and Stockton using the K-S test for goodness of fit. The analyses were performed using the statistical package, STATGRAPHICS (Statistical Graphics Corporation, Ver 5.0). The concentration data were also fitted on an annual basis (2002 was neglected due to insufficient data points), and seasonal basis in order to test any year-to-year variations, and seasonal variations in the pdf parameters. Summer and winter were described previously. For the entire sampling interval, the lognormal distributions are satisfactorily fitted to the measured concentrations (Table 8). At Potsdam, the year-to-year differences in μ and σ were small compared to the seasonal variations. This observation indicates that the overall effect of variations in sources, source strengths, and meteorology on an annual basis were comparable among the years. The observations made at Stockton were similar, although during any given period, the geometric means at Stockton were greater than those at Potsdam. At Stockton, the σ in 2005 was much larger than the other years. However, this observation must be interpreted carefully since the sample size during this period was much smaller (only 22) compared to the other periods. A comparison of the pdf parameters between the seasons at Potsdam, and at Stockton indicate that the geometric mean and standard deviation were greater in the summers compared to winters. This observation supports the hypothesis that there is an overall increase in the rate of secondary formation of fine PM during the summers. Differences in the meteorological conditions between the seasons, and between locations may also contribute to differences in the width of σ between the seasons, and locations, respectively.
Conclusions In general, the monthly average mass concentration time series at both locations exhibited a seasonal behavior with pronounced summer annual peaks.
365
There were also indications of the presence of a winter peak. It was hypothesized that the summer peak was related to the overall increase in the secondary aerosol formation rates, resulting in increased concentrations of secondary sulfate and SOA in the particle phase, while the winter peak was related to increased ammonium nitrate partitioning to the particle phase, wood burning for residential heating, and condensation of semi-volatile organic species. An ion balance calculation using the concen trations of major inorganic ions (SO2 4 , NO3 and NHþ 4 ) indicated that the acidity at Stockton during each season was mostly higher than the corresponding acidity at Potsdam. This observation was attributed to the fact that Potsdam has larger ammonia sources compared to Stockton because of the large number of dairy farms located near Potsdam. Lognormal frequency distribution functions were fitted to fine PM mass concentrations at both locations on an annual and seasonal basis. It was proposed that the increase in the geometric mean (μ), and geometric standard deviation (σ) of the fitted distributions during the summers reflected the increased overall rate of secondary aerosol formation during the summers compared to the winters at both locations. Acknowledgments The authors wish to thank Dr. Micheal Milligan at SUNY, Fredonia, NY for sampling at Stockton, Dr. Soon Onn Lai, and Dr. Chris Brancewicz for sampling at Potsdam. This work was supported by New York State Energy Development Authority (NYSERDA), through contract numbers 6083 and 7919. Although this work was funded by NYSERDA, it does not necessarily reflect the views of the agency and no official endorsement should be inferred.
References Draxler, R. R., & Rolph, G. D. (2003). HYSPLIT (Hybrid Single-Particle Lagrangian Integrated Trajectory) Model. NOAA Air Resources Laboratory, Silver Spring, MD. Access via NOAA ARL READY Website http://www.arl. noaa.gov/ready/hysplit4.html. deNevers, N., Lee, K. W., & Frank, N. H. (1979). Patterns in TSP distribution. Journal of the Air Pollution Control Association, 29, 32–35. Hopke, P. K., Liu, W., Han, Y. J., Yi, S. M., Holsen, T. M., Scott, C., & Milligan, M. (2003). Measured summertime concentrations of particulate components, Hg0, and speciated polycyclic aromatic hydrocarbons at rural sites in New York State. Environmental Pollution, 123, 413–425. Kao, A. S., & Friedlander, S. K. (1994). Chemical signatures of the Los Angeles aerosol (dp<3.5 mm). Aerosol Science and Technology, 21, 283–293.
366 Kao, A. S., & Friedlander, S. K. (1995). Frequency distributions of PM10 chemical components and their sources. Environmental Science & Technology, 29, 19–28. Karaca, F., Alagha, O., & Erturk, F. (2005). Statistical characterization of atmospheric PM10 and PM2.5 concentrations at a non-impacted suburban site of Istanbul, Turkey. Chemosphere, 59, 1183–1190. Liu, W., Hopke, P. K., Han, Y.-J., Yi, S.-M., Holsen, T. M., & Cybart, S., et al. (2003b). Application of receptor modeling to atmospheric constituents at Potsdam and Stockton, NY. Atmospheric Environment, 37, 4997–5007. Liu, W., Hopke, P. K., & VanCuren, R. A. (2003a). Origins of fine aerosol mass in the western United States using positive matrix factorization. Journal of Geophysical Research, 108(D23), 4716. Lu, H.-C. (2002). The statistical characters of PM10 concentration in Taiwan area. Atmospheric Environment, 36, 491–502. McMurry, P. H. (2000). A review of atmospheric aerosol measurements. Atmospheric Environment, 34, 1959–1999. Natural Resources Canada, Canadian Forest Services (accessed on 07/28/2006) http://www.nrcan.gc.ca/cfs-scf/science/ prodserv/firereport/archives_e.php Ott, W. (1990). A physical explanation of the log normality of pollutant concentrations. Journal of Air & Waste Management Association, 40, 1378–1383. Paatero, P., & Hopke, P. K. (2003). Discarding and downweighting high noise variable in factor analysis models. Analytica Chimica Acta, 490, 277–289.
Environ Monit Assess (2008) 144:351–366 Research Triangle Institute (2003a). Standard operating procedure for PM2.5 anion analysis, environmental & industrial sciences division, research triangle institute, Research triangle Park, NC’ (accessed on 04/23/2006). http://www. epa.gov/ttn/amtic/files/ambient/pm25/spec/anionsop.pdf. Research Triangle Institute (2003b). Standard operating procedure for PM2.5 cation analysis, environmental & industrial sciences division, research triangle institute, Research triangle Park, NC’ (accessed on 04/23/2006). http://www. epa.gov/ttn/amtic/files/ambient/pm25/spec/cationsop.pdf. Schwab, J. J., Felton, H. D., & Demerjian, K. L. (2004). Aerosol chemical composition in New York State from integrated filter samples: Urban/rural and seasonal contrasts. Journal of Geophysical Research, 109, D16S05. Seinfeld, J. H., & Pandis, S. N. (1998). Atmospheric chemistry and physics. New York: Wiley. Sunder Raman, R. (2006). Investigation of the nature and behavior of carbonaceous species and source apportionment of ambient particulate matter. Dissertation submitted to the Department of Chemical and Biomolecular Engineering, Clarkson University, Potsdam, NY. Sunder Raman, R., & Hopke, P. K. (2006). An ion chromatographic analysis of water-soluble, short-chain organic acids in ambient particulate matter. International Journal of Environmental Analytical Chemistry, 86(10, 20), 767–777 (Note: First author last name misspelt in the journal as Sundar Raman). Sunder Raman, R., & Hopke, P. K. (2007). Carbonaceous aerosol at two rural locations in New York State: Characterization and behavior. Journal of Geophysical Research (in press).