You are on page 1of 9

Environmental Science & Technology | 3b2 | ver.

9 | 17/3/011 | 12:20 | Msc: es-2010-03831w | TEID: cdm00 | BATID: 00000 | Pages: 8.36

ARTICLE

pubs.acs.org/est

1 Heterogeneity of Coarse Particles in an Urban Area


2 Uma Ramesh K. Lagudu,† Suresh Raja,† Philip K. Hopke,†,* David C. Chalupa,‡
3 Mark J. Utell,‡ Gary Casuccio,§ Traci L. Lersch,§ and Roger R. West§

4 Center for Air Resource Engineering and Science, Clarkson University, Potsdam, New York 13699, United States

5 Departments of Medicine and Environmental Medicine, University of Rochester Medical Center, Rochester, New York 14642, United
6 States
§
7 RJ Lee Group, Monroeville, Pennsylvania 15146, United States
8 bS Supporting Information
9 ABSTRACT: The variation in composition and concentration of
10 coarse particles in Rochester, a medium-sized city in western
11 New York, was studied using UNC passive samplers and
12 computer-controlled scanning electron microscopy (CCSEM).
13 The samplers were deployed in a 5  5 grid (2 km  2 km per
14 grid cell) for 23 week periods in two seasons (September 2008
15 and May 2009) at 25 different sites across Rochester. CCSEM
16 analysis yielded size and elemental composition for individual
17 particles and analyzed more than 800 coarse particles per sample.
18 Based on the composition as reflected in the fluoresced X-ray
19 spectrum, the particles were grouped into classes with similar
20 chemical compositions using an adaptive resonance theory
21 (ART) network. The mass fractions of particles in the identified
22 classes were then be used to assess the homogeneity of composi-
23 tion and concentration across the measurement domain. These
24 results illustrate how particle sampling using the UNC passive
25 sampler coupled with CCSEM/ART can be used to determine
26 the concentration and source of the coarse particulate matter at multiple sites. The particle compositions were dominated by
27 elements suggesting that the major particle sources are road dust and biological particles. Considerable heterogeneity in both
28 composition and concentration were observed between adjacent sites as indicated by cofficient of divergence analyses.

30 ’ INTRODUCTION concentrations were more homogeneous. Therefore, to assess 50

31 Particulate matter (PM) exposures have drawn considerable community level exposures, large numbers of samplers at multi- 51

32 attention because of links to human mortality and morbidity.13 ple locations are required.10 Hwang et al.14 highlighted significant 52

33 Size is one of the important parameters that determine the potential for exposure misclassification in time series epidemiologic 53

34 environmental and health effects of aerosol particles along with studies when regressing health outcomes against source contribu- 54

35 concentration, structure, and chemical composition. Relationships tions estimated at a single central monitoring site. Conventional 55

36 between particles and adverse health effects including mortality active samplers used to assess area level particulate exposures are 56

37 have been observed even at low levels of exposure.4,5 costly and labor intensive to obtain acceptable spatial resolution. 57

38 The health effects of coarse particles need to be considered diff- Passive samplers are available that are inexpensive and easier 58

39 erently from those of the fine fraction since they deposit differently in to use than conventional samplers. Therefore, a large number of 59

40 the respiratory system and may produce different health effects samplers can be deployed yielding more representative spatial 60

41 compared to fine particles.6 The type and severity of health effects measurements with better measures of precision (duplicate sam- 61

42 depend on the particle size, concentration, and composition.7 Clinical ples). Ott et al.1517 have shown passive sampling to be a feasible 62

43 studies8,9 have shown that acute exposure to coarse particles can method to sample coarse particulate matter in an urban area. 63

44 produce cardiopulmonary changes. Passive particulate samplers are intended to monitor ambient, 64

45 Because of their relatively short atmospheric lifetime, previous


46 studies have shown that coarse particles are spatially heteroge- Received: November 15, 2010
47 neous.1012 Burton et al.13 collected PM samples from eight sites Accepted: March 10, 2011
48 in a metropolitan area in Philadelphia and showed that the Revised: March 4, 2011
49 coarse PM concentrations varied spatially, while PM2.5 and PM10

r XXXX American Chemical Society A dx.doi.org/10.1021/es103831w | Environ. Sci. Technol. XXXX, XXX, 000–000
Environmental Science & Technology ARTICLE

Figure 1. Map showing locations of the sampling sites and relationship of sampling domain to New York state.

65 indoor, or occupational exposure over a period of hours to weeks. other similar structures at heights of approxmately 3 m above the 94
66 They have the potential to be used as an area monitor or as a ground in a 5  5 grid (2 km  2 km per grid cell) for 23 week 95
67 personal sampler.18,19 Since particles deposit on the sampler by periods in two seasons (September 2008 and May 2009). Locations 96
68 diffusion and gravity, longer sampling times are required but were selected to minimize the impact of local vegetation (trees) so 97
69 provide assessments of long-term average exposures. that the surroundings were as open as possible. 98
70 To date, very little information has been available to characterize During the first campaign in September, samplers were placed 99
71 the spatial variation of PM size and composition on a local scale. on September 17, 2008 and removed on October 1, 2008. During 100
72 Previous studies have explored intracity variations of PM concen- the spring, the samplers were placed on April 30, 2009 and removed 101
73 tration, size, and composition of airborne particles.20,21 However, on May 21, 2009. Figure 1 shows the locations of the samplers 102 F1
74 these studies do not give relevant information on local variability, for installed in Rochester. 103
75 example in an urban area, in order to be able to assess epidemio- The UNC passive sampler consists of a standard scanning electron 104
76 logical exposure of PM. microscope (SEM) stub, a collection substrate, and a protective 105
77 In this work, UNC passive samplers2224 were deployed at 25 mesh cap. During sampling, particles are transported by gravity and 106
78 locations in Rochester, NY to collect integrated particle samples diffusion through the 150-μm-diameter holes in the mesh cap and 107
79 (particle size between 0.05 and 15 μm). Computer-controlled deposit on a substrate mounted on the stub. The stub is oriented 108
80 scanning electron microscopy (CCSEM) was used to measure such that the substrate is horizontal. After sampling, the mesh cap is 109
81 individual particle size, shape, and elemental composition of the removed, the stub is placed in a scanning electron microscope (Aspex 110
82 coarse PM collected on the samplers. Adaptive Resonance Theory Personal SEM), and the particles are counted, sized, and composition 111
83 (ART) neural networks were then used to classify the measured assayed using CCSEM analytical procedures. 112
84 particle composition into homogeneous groups. Methods to de- CCSEM. PM samples collected on the sampler substrate were 113
85 termine the spatial variation in PM concentration, size, and comp- analyzed using CCSEM.25,26 Briefly, CCSEM scans the collec- 114
86 osition are presented and discussed. tion substrate of the SEM stub for individual particles and pr- 115
ovides a fluoresced X-ray spectrum and an image of each particle. 116

87 ’ EXPERIMENTAL METHODS The analysis is performed by rastering the electron beam over the 117
sample while monitoring the resultant backscattered signal. At 118
88 Sample Collection. The UNC Passive Sampler was devel- each point, the image intensity is compared to a preset threshold 119
89 oped by Wagner and Leith.2224 The passive sampler shelter level. Once a coordinate is reached where the signal is above the 120
90 design described by Ott et al.16 was used during this campaign to threshold level, the electron beam is driven across the particle in 121
91 avoid any effects of precipitation getting into the samplers. a preset pattern to determine the size of the particle. Upon 122
92 The samplers were deployed at 25 locations in the Rochester, measurement of the particle size, the elemental composition of 123
93 NY metropolitan area. They were placed on telephone poles and the particle is determined through collection of characteristic 124

B dx.doi.org/10.1021/es103831w |Environ. Sci. Technol. XXXX, XXX, 000–000


Environmental Science & Technology ARTICLE

Figure 2. Spatial variation of coarse PM mass concentrations (ng/m3) in Rochester, NY. Solid dots represent the sampling locations.

Figure 3. Wind roses for the fall 2008 (left) and spring 2009 (right) sampling periods.

125 X-rays using energy dispersive spectroscopy (EDS) techniques. observed elements include carbon (C), sodium (Na), magne- 136
126 Individual particles characterized during the CCSEM analysis sium (Mg), aluminum (Al), silicon (Si), phosphorus (P), sulfur 137
127 are grouped into particle type classes based on their elemental (S), chlorine (Cl), potassium (K), calcium (Ca), titanium (Ti), 138
128 composition. The mass of an individual particle was calculated chromium (Cr), manganese (Mn), iron (Fe), nickel (Ni), copper 139
129 by multiplying the assigned density of the particle by its volume. (Cu), zinc (Zn), barium (Ba), and lead (Pb). 140
130 Each particle was assigned a density based on a common oxide Sampling and Analysis Precision. Replicate samplers were 141
131 in proportion to the elements present as determined by the EDS deployed at 5 locations during each sampling campaign to assess 142
132 analysis.25,26 Because of the small mass being analyzed and the reproducibility of the mass measurements. The resulting data 143
133 the relatively short accumulation time for each spectrum, only have been separated by size into μm intervals and the results are 144
134 a limited suite of elements are observed after application of the presented in Figure S1 in the Supporting Information. There is 145
135 noise reduction procedure described in the next section. The a moderate correlation between the replicate samples, but the 146

C dx.doi.org/10.1021/es103831w |Environ. Sci. Technol. XXXX, XXX, 000–000


Environmental Science & Technology ARTICLE

Table 1. Particle Classes Obtained from Clustering of Particle Data Obtained from Two Seasons Together
mean fall mean spring
cluster number particle type source type concentration (ng/m3) concentration (ng/m3)

1 Ca road dust 115.7 30.1


2 CP (negligible P) biological 55.4 27.3
3 CCaZn (High C) C road dust 6.8 7.4
4 NaMgCa (High CaMg) salt 13.6 89.3
5 CMgAlSiKMnFe (CSiAlFeMg) C road dust 7.6 7.0
6 PCa (Ca) biological 3.1 105.6
7 STiMnFeCuBaPb (Fe) road dust 4.5 0.5
8 MgAlSiKMnFe (SiAl) road dust 62.0 16.1
9 AlSiK (SiAl) road dust 19.2 16.8
10 AlSiK (SiAl) More Si, Al road dust 98.0 66.7
11 CPKNi (C) C road dust 0.4 12.5
12 Si road dust 77.6 1.7
13 AlSiKMn (SiAl) road dust 11.5 154.9
14 CSiCr (CSi) C road dust 2.5 25.9
15 NaMgCa (CaMgNa) salt 163.2 13.6
16 CrMnFeNiCuZnBaPb (Fe) Cr-rich 43.8 2.8
17 CAlSiTiMnFeNiCuZnBaPb (CSiAl) C road dust 0.4 27.2
18 CAlSiK-Mn (SiCAl) C road dust 24.4 24.9
19 MgAlSiSKMnPb (SiAl) road dust 208.5 61.2
20 CAlZn (CAl) C road dust 12.2 25.3
21 CAlSiK (CSiAl) C road dust 9.5 24.8
22 Si road dust 157.3 68.4
23 CP biological 4.7 59.1
24 CPCrZn (C) biological 62.8 0.3
25 CTiCrFeNiCuZnBaPb (FeC) C road dust 2.4 7.7
26 CrFeBa (Fe) Cr-rich 3.5 16.2

147 limited number of particles in any given size interval reduced the particles at a single site was estimated by summing the concentration 175
148 correlation among the results. contribution of all the particles at that site. These data were used 176
149 Data Analysis. CCSEM measurements provide a representa- finding the spatial variation of the coarse particles over the sampling 177
150 tion of the particle composition, but can rarely be used to provide locations. The quality of these results has been assessed using 178
151 accurate elemental concentrations on a particle-by-particle basis. the replicate samples. The analysis of the replicate sample data is 179
152 Therefore, in order to use the semiquantitative data to obtain new presented in the Supporting Information. For those sites with 180
153 quantitative variables, particles need to be classified into homo- replicate samples, the average value of the two samples was used. 181
154 geneous groups. Particle classes or groups represent the types of There was generally good agreement among the replicate samples. 182
155 particles present in the air and the mass of particles in a given class is Application of ART-2a Model. The chemical information 183
156 a quantitative measure of particle composition.27 derived from CCSEM serves as a basis for classification of particles 184
157 Because there are peaks in the spectrum that arise from statistical into classes with similar composition. The normalized unit length 185
158 fluctuations in the X-ray spectrum arising from the short acquistion data were analyzed with an adaptive resonance theory (ART-2A) 186
159 time (5 s/particle) and influence of the background substrate, the algorithm.31 This dynamic classification system has been used 187
160 data were subjected to noise reduction and composition data were previously for classification of particles collected on filter samples.32 188
161 normalized to unit length. Detailed steps for pretreatment of single In the present work, particle classification using ART-2a was per- 189
162 particle data obtained from CCSEM measurements have been given formed for different vigilance factors, and a vigilance factor of 0.5 190
163 by Kim et al.28 Any X-ray count for a given element less than twice was found to yield a good classification with homogeneous classes in 191
164 the square root of total X-ray count for an individual particle was set terms of the chemical elements contained in each group. The ART- 192
165 to zero. After noise reduction, the X-ray spectrum for each particle 2a algorithm was run with a vigilance factor of 0.5 and a learning rate 193
166 data was normalized to unit length. The volume was estimated of 0.5 for all of the samples from both seasons (fall and spring). 194
167 from the maximum diameter (dmax) and perpendicular diameter The numbers of measured particles for the fall and spring season 195
168 (dperp).29 The mass of each particle was then estimated from the were 21 596 and 29 000, respectively. Particles with diameters less 196
169 volume of a spheroid of revolution and an estimated density based than 2.5 μm and above 10 μm were discarded, and the remaining 197
170 on the elemental composition. The concentration of each particle “coarse” particles are the subject of this paper. The fall and spring data 198
171 has been estimated with the deposition velocity model discussed by sets were combined with a total coarse particle number of 11 550. The 199
172 Ott et al.16 using a surface roughness length of 0.530 to determine the ART-2a algorithm was run using a standard UNIX network cluster. 200
173 friction velocity. The diffusivity of the particle was estimated using Only resulting particle groups/clusters representing more than 1% of 201
174 the StokesEinstein equation. The total concentration of all coarse the total number of particles were considered for further analysis. 202

D dx.doi.org/10.1021/es103831w |Environ. Sci. Technol. XXXX, XXX, 000–000


Environmental Science & Technology ARTICLE

Figure 4. Spatial variation of mass concentration (ng/m3) of combined soil-related clusters during fall (left) and spring (right). Solid dots represent
sampling locations.

Figure 5. Spatial variation of mass concentration (ng/m3) of combined carbon/soil-related clusters during fall (left) and spring (right). Solid dots
represent sampling locations.

203 Each cluster obtained from the ART-2a analysis had elemental compared to the elemental average of all the coarse particles 207
204 information, aerodynamic diameter, and mass of each particle. for that season. If a particular element was found to have a higher 208
205 The masses were summed to yield the total mass in each cluster. average value than the overall elemental average, it was consid- 209
206 The elemental average of each of the elements in each cluster was ered to be a defining element in the class. In this way all the 210

E dx.doi.org/10.1021/es103831w |Environ. Sci. Technol. XXXX, XXX, 000–000


Environmental Science & Technology ARTICLE

Figure 6. Spatial variation of mass concentration (ng/m3) of combined biological clusters during fall (left) and spring (right). Solid dots represent
sampling locations.

211 classes of the clusters have been identified. The spatial variations similar in their elemental content (e.g., clusters 21 and 22 in 242
212 have been plotted using a spline interpolation scheme (ArcGIS Table 1). However, these clusters show differences in the ratios 243
213 V9.2). Although various interpolation schemes such as kriging of the constituent elements. These clusters are likely to arise from 244
214 and inverse distance weighting, etc. were tried, spline interpola- the same source type. 245
215 tion provided smoother and more realistic plots. From Table 1, high concentrations of various species, e.g., Ca, 246
C, CaMg, and SiAl, can also be seen in a given cluster. SiAl 247

216 ’ RESULTS AND DISCUSSION generally indicates crustal material. In general, there is quite limited 248
direct suspension of soil particles because of the limited amount of 249
F2 217 Spatial Variation of Coarse Particle Mass. Figure 2 shows open space in urban areas that is not covered with vegetation. 250
218 the spatial variation in the mass concentrations (ng/m3) of However, soil and other materials deposited on road surfaces are 251
219 coarse PM across the 25 sites in Rochester during the fall-08 resuspended by traffic. Thus, particle types such as in Clusters 1, 252
220 and the spring-09 sampling campaigns. The wind roses for these 710, 12, 13, 19, and 22 are likely to be road dust/soil. 253
F3 221 two sampling periods are shown in Figure 3. These plots present The presence of carbon along with these elements can still 254
222 the spatial variation across the sampling domain. The spatial indicate that such particle types are road dust but with the addition 255
223 distributions of coarse PM in Rochester were highly variable of carbonaceous material. Road abrasion of asphalt roads or tire 256
224 with concentrations ranging from 100 to 3230 ng/m3 in the fall wear deposited on road surfaces can be sources of carbon in 257
225 and 410 to 2050 ng/m3 in the spring (Supporting Information addition to tailpipe emissions or deposition of primary biological 258
226 Table S1). During the fall 2008 period, the wind direction was materials. Particle types where carbon was found to be associated 259
227 predominately from the southwest. with elements like Zn suggests that such coarse particle types 260
228 It can be seen from the fall plot (left panel Figure 2) that there are derived from combusted lubricating oil and/or tire wear.3337 261
229 is a high concentration in the southeastern corner of the sampling Ba and Cu are observed in several clusters and suggest brake wear 262
230 domain. This site is adjacent to an agricultural field so that har- particles.38 Thus, there is a second group of road dust clusters incl- 263
231 vesting likely contributed to the higher observed concentrations. For uding cluster numbers 3, 5, 14, 17, 18, 20, 21, and 25 that include 264
232 the spring samples (right panel), high mass concentrations are crustal elements plus high concentrations of carbon. 265
233 observed in the northwestern corner of the sampling domain. This There are several clusters with carbon and/or phosphorus that 266
234 area is close to a major industrial zone in Rochester. Wind directions may represent biologically derived particles.39 These clusters are 267
235 during this late April to May period were predominately westerly. 2, 6, 11, 23, and 24. Because of the constraints on this project, 268
236 Particle Class Membership. Clustering of the coarse particles images are not available for the individual particles so we cannot 269
237 yielded 26 different particle types with mass values greater than confirm these assignments based on morphology. Future work is 270
T1 238 1% of the total estimated mass. Table 1 presents the defined planned that will also acquire images so confirmatory studies can 271
239 particle classes and the mass concentrations for each season. The be performed using individual particle morphology. 272
240 particle types are dominated by the presence of silicon, calcium, There are two classes that are high in sodium suggesting a 273
241 magnesium. and carbon. Some particle types appear to be very residue of salt that is used on Rochester roads during the winter 274

F dx.doi.org/10.1021/es103831w |Environ. Sci. Technol. XXXX, XXX, 000–000


Environmental Science & Technology ARTICLE

275 months. This group includes cluster numbers 5 and 15. Finally, Table 2. Maximum and Minimum Coefficient of Divergence,
276 two clusters (Nos. 10 and 26) had measurable chromium that Fall and Spring, Rochester
277 could derive from yellow road paint that is commonly used on
site # max min avg site # max min avg
278 roads, but also could arise from particles being worn from the
279 main line railroad40 that runs through the sampling domain and S#2243 0.870 0.420 0.597 S#1580 0.936 0.555 0.629
280 from an industrial area that is in the upper left corner of the S#2244 0.806 0.365 0.549 S#1582 0.910 0.547 0.648
281 sampled domain. S#2246 0.829 0.474 0.601 S#1585 0.913 0.481 0.571
282 Spatial Variation of Particle Types. To examine the spatial S#2247 0.853 0.406 0.605 S#1586 0.891 0.482 0.566
283 distributions of these groups of clusters that likely derive from the S#2248 0.831 0.424 0.561 S#1587 0.897 0.439 0.539
284 same underlying sources, the mass contributions for each cluster
S#2250 0.816 0.412 0.547 S#1588 0.867 0.513 0.607
285 have been summed and the resulting mass concentrations of the
S#2251 0.847 0.441 0.569 S#1592 0.870 0.557 0.659
286 road dust, high carbon road dust, and biological particles are
F4 287 plotted in Figures 4 to 6. The spatial variation of the Cr-rich and S#2252 0.809 0.365 0.540 S#1598 0.868 0.506 0.615
F6 288 salt groups are shown in Supporting Information Figures S2 and S#2253 0.821 0.402 0.570 S#1604 0.877 0.481 0.606
289 S3, respectively. The spatial variations of each of the identified S#2254 0.922 0.640 0.775 S#1609 0.893 0.519 0.624
290 particle types are provided in the Supporting Information for this S#2255 0.812 0.395 0.577 S#1618 0.870 0.494 0.591
291 paper as Figures S4 to S30. S#2256 0.834 0.480 0.588 S#1620 0.908 0.563 0.634
292 It can be seen from Figure 4 that the spatial patterns are very S#2258 0.835 0.479 0.602 S#1648 0.879 0.477 0.588
293 similar to those seen for the total particle mass shown in Figure 2. S#2259 0.878 0.505 0.646 S#1666 0.871 0.481 0.554
294 This crustal element dominated group of clusters represents 64% S#2260 0.855 0.461 0.570 S#1671 0.917 0.516 0.625
295 of the fall particle mass and 47% of the spring mass. As mentioned S#2261 0.857 0.536 0.619 S#1688 0.890 0.534 0.605
296 above, there is an agricultural area in the southeastern corner of
S#2262 0.840 0.462 0.571 S#1707 0.903 0.541 0.649
297 the domain that was actively harvesting vegetables during the fall
S#2263 0.808 0.395 0.585 S#1712 0.942 0.509 0.607
298 sampling period. Winds during the fall were predominately from
299 the southwest (Figure 3) bringing particles into the sampling S#2265 0.868 0.406 0.594 S#1718 0.867 0.482 0.593
300 domain from the south. In the spring samples, Figure 4 (right), S#2266 0.843 0.513 0.611 S#1721 0.927 0.439 0.582
301 there appear to be high concentrations in the vicinity of I-590 in S#2267 0.839 0.476 0.567 S#1722 0.870 0.511 0.595
302 the southwest corner of the domain. Winds in the spring were S#2269 0.864 0.424 0.586 S#1737 0.846 0.480 0.586
303 stronger and more westerly in direction bringing particles from S#2270 0.942 0.640 0.758 S#1745 0.890 0.514 0.613
304 the industrial area in the northwestern corner. For the high- S#2271 0.851 0.475 0.633 S#1747 0.852 0.480 0.578
305 carbon road dust particle types, the areas of high values are
306 generally clustered around the major roads shown in Figure 5. pattern that suggests a road source, while the other plot (Figure S29) 337
307 These particle types contributed similar fractions of the coarse appears to be related to the railroad. 338
308 aerosol mass (11%) in both seasons. Because the carbon is being Coefficient of Divergence (COD). There is a visual degree of 339
309 attributed to tire wear and other motor vehicle related material, heterogeneity observable in the plots described in the prior 340
310 the spatial patterns are logical. The carbonaceous soil also section. However, to provide a quantitative measure of spatial 341
311 appeared to be emitted by the agricultural area. The biological heterogeneity, the coefficient of divergence (COD)41 was calcu- 342
312 particles (Figure 6) showed a very different spatial pattern with lated using the particle type concentrations at the sampling sites. 343
313 higher values in the residential areas away from the highways. The COD is defined as 344
This group contributed about 3 times more mass in the spring vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
!ffi
314
than in the fall suggesting pollen as a major portion of the u
315 u1 n xij  xik 2
316 biological material. Biological material represented 6% of the CODjk ¼ t ∑
317 fall coarse particle mass and 17% of the average spring mass n i ¼ 1 xij þ xik
318 concentration.
319 The sodium-dominated particle types (Figure S2) show high- where xij is the source contribution per sampling interval, i, 345
320 er contributions in the fall than the spring. The other elements in estimated at site j, j and k represent two sites, and n is the number 346
321 these clusters are Ca and Mg. For the spring samples, the pattern of sampling intervals. The COD provides a measure of the degree 347
322 of high concentrations is strongly related to the I-590/I-490 area. of uniformity among sampling sites. As the estimated particle 348
323 Thus, there appears to be some residual from the winter salt concentrations at two sites become more similar, the COD app- 349
324 application to the roads. However, the fall pattern is less clear roaches zero. Alternatively, as the estimated concentrations 350
325 with the high values observed in three corners of the domain and diverge, the COD approaches one. The maximum, minimum, 351
326 the highest values again in the southeastern agricultural area. and average values of all the coefficients of divergence (COD) are 352
327 Examining the meteorological data for the sampling period, there given in Table 2. All of the COD values are presented in SI Tables 353 T2
328 were no periods of snow or freezing precipitation. There could be S2 and S3. The average values indicate that the sites are rather 354
329 application of soil amendments during the fall (liming the soil dissimilar during both seasons. The sites are relatively better 355
330 to control pH for example), but the exact nature of the fall Na correlated in the spring compared to the fall. 356
331 particles is unknown. During the spring, the COD values among the sites were 357
332 The spatial patterns of the Cr-rich particles are shown in higher and did not exhibit any specific trends showing that the 358
333 Figure S3. For the fall samples, there is a high area at the southern sites are more independent of each other. The value of the 359
334 edge of the domain and in the northeastern corner. In the spring maximum COD is always above 0.7, greater than the criterion 360
335 samples, the high areas appear to be downwind of the railroad. suggested by the USEPA to indicate heterogeneity (COD > 361
336 Examination of one individual cluster plot (Figures S19) shows a 0.2).42 High values of COD are found in the present study with a 362

G dx.doi.org/10.1021/es103831w |Environ. Sci. Technol. XXXX, XXX, 000–000


Environmental Science & Technology ARTICLE

363 minimum value of 0.365. Lower values might be expected when (6) Sandstrom, T.; Nowak, D.; Van Bree, L. Health effects of coarse 418
364 sampling sites are only 2 km apart and the averaging times are particles in ambient air: messages for research and decision-making. Eur. 419
365 21 days to 30 days. Thus, there is very high spatial variability for Respir. J. 2005, 26, 187–188. 420
366 coarse particles across this urban area. (7) Hinds, W. C. Aerosol Technology: Properties, Behavior, and 421
367 Further analyses of these data could be pursued using the Measurement of Airborne Particles; J. Wiley & Sons, Inc.: New York, 1999. 422
(8) Graff, D. W.; Cascio, W. E.; Rappold, A.; Zhou, H.; Huang, 423
368 spatial variation in a factor analysis approach such as RMAPS43 or Y. C. T.; Devlin, R. B. Exposure to concentrated coarse air pollution 424
369 that described by Paatero et al.44 These approaches will be particles causes mild cardiopulmonary effects in healthy young adults. 425
370 explored on a larger data set for which such analyses may be Environ. Health Perspect. 2009, 117, 1089–1094. 426
371 more applicable, but they should permit quantitative identifica- (9) Huang, Y. C.; Ghio, A. J. Vascular effects of ambient pollutant 427
372 tion and apportionment of the major coarse particle sources. particles and metals. Curr. Vasc. Pharmacol. 2006, 4, 199–203. 428
373 This work demonstrates the utility of passive sampling coupled (10) Jerrett, M.; Burnett, R. T.; Ma, R. J.; Pope, C. A.; Krewski, D.; 429
374 with CCSEM/ART to determine the concentration and composi- Newbold, K. B.; Thurston, G.; Shi, Y,L; Finkelstein, N.; Calle, E. E.; 430
375 tion of the coarse particulate matter at multiple sites and hence, to Thun, M. J. Spatial analysis of air pollution and mortality in Los Angeles. 431
376 assess the spatial variability. The spatial variation plots showed Epidemiology 2005, 16, 727–736. 432
377 the temporal (spring vs fall) as well as the spatial variability to be (11) Monn, C. Exposure assessment of air pollutants: a review on 433
spatial heterogeneity and indoor/outdoor/personal exposure to sus- 434
378 significant. These results illustrate that there could be significant pended particulate matter, nitrogen dioxide and ozone. Atmos. Environ. 435
379 exposure misclassification in epidemiological studies that depend on 2001, 35, 1–32. 436
380 data from one or a few central monitoring stations. However, the (12) Wilson, W. E.; Suh, H. H. Fine particles and coarse particles: 437
381 passive monitor method could be used to assess exposure to coarse concentration realtionships relevant to epidemiological studies. J. Air 438
382 PM at a local scale. The resulting concentration estimates could then Waste Manage. Assoc. 1997, 47, 1238–1249. 439
383 be incorporated into epidemiological studies to better assess the (13) Burton, R. M.; Suh, H. H.; Koutrakis, P. Spatial variation in 440
384 association between health effects and exposure to coarse particles. particulate concentrations within metropolitan Philadelphia. Environ. 441
Sci. Technol. 1996, 30, 400–407. 442
(14) Hwang, I.; Hopke, P. K.; Pinto, J. P. Source apportionment and 443
385 ’ ASSOCIATED CONTENT spatial distributions of coarse particles during the Regional Air Pollution 444
Study. Environ. Sci. Technol. 2008, 42, 3524–3530. 445
386 bS Supporting Information. This material is available free (15) Ott, D. K.; Cyrs, W.; Peters, T. M. Passive measurement of 446
387 of charge via the Internet at http://pubs.acs.org. coarse particulate matter, PM102.5. J. Aerosol Sci. 2008, 39, 156–167. 447
(16) Ott, D. K.; Peters, T. M. A shelter to protect a passive sampler 448
388 ’ AUTHOR INFORMATION for coarse particulate matter, PM102.5. Aerosol Sci. Technol. 2008, 449
42, 299–309. 450
389 Corresponding Author (17) Ott, D. K.; Kumar, N.; Peters, T. M. Passive sampling to capture 451
390 *Fax: (315) 268 4410; e-mail: hopkepk@clarkson.edu. spatial variability in PM102.5. Atmos. Environ. 2008, 42, 746–756. 452
(18) Yamamoto, N.; et al. A passive sampler for airborne coarse 453
particles. J. Aerosol Sci. 2006, 37, 1442–1454. 454
(19) Vinzents, P. S. A passive personal dust monitor. Ann. Occup. 455
391 ’ ACKNOWLEDGMENT Hyg. 1996, 40, 261–280. 456
392 This work was supported by U.S. Environmental Protection (20) Henderson, S. B.; Beckerman, B.; Jerrett, M.; Brauer, M. 457
393 Agency’s Science to Achieve Results (STAR) Program through a Application of land use regression to estimate long-term concentrations 458
394 subcontract from the University of Rochester PM and Health of traffic related nitrogen oxides and fine particulate matter. Environ. Sci. 459
395 Center Grant RD832415. Although the research described in this Technol. 2007, 41, 2422–2428. 460
(21) Jerrett, M.; Arain, M. A.; Kanaroglou, P.; Beckerman, B.; 461
396 article has been funded wholly or in part by the United States
Crouse, D.; Gilbert, N. L.; Brook, J. R.; Finkelstein, N.; Finkelstein, 462
397 Environmental Protection Agency, it has not been subjected to M. M. Modeling the intraurban variability of ambient traffic pollution in 463
398 the Agency’s required peer and policy review and therefore, does Toronto, Canada. J. Toxicol. Environ. Health, Part A 2007, 70, 200–212. 464
399 not necessarily reflect the views of the Agency and no official (22) Wagner, J.; Leith, D. Passive aerosol sampler. I: Principle of 465
400 endorsement should be inferred. operation. Aerosol Sci. Technol. 2001, 34, 186–192. 466
(23) Wagner, J.; Leith, D. Passive aerosol sampler. II: Wind tunnel 467
experiments. Aerosol Sci. Technol. 2001, 34, 193–201. 468
401 ’ REFERENCES (24) Wagner, J.; Leith, D. Field tests of a passive aerosol sampler. 469
402 (1) Brunekreef, B.; Holgate, S. T. Air pollution and health. Lancet J. Aerosol Sci. 2001, 32, 33–48. 470
403 2002, 360, 1233–1242. (25) Hopke, P. K., Casuccio, G. S. Scanning Electron Spectroscopy. 471
404 (2) Stenfors, N.; Nordenhall, C.; Salvi, S. S.; Mudway, I.; Soderberg, In Receptor Modeling for Air Quality Management; Hopke, P.K., Ed.; 472
405 M.; Blomberg, A.; Helleday, R.; Levinz, J.-O.; Holgate, S. T.; Kelly, F. J.; Elsevier Science: Amsterdam, 1991; pp 149212. 473
406 Frew, A. J.; Sandstrom, T. Different airway inflammatory responses in (26) Mamane, Y.; Willis, R.; Conner, T. Evaluation of computer- 474
407 asthmatic and healthy humans exposed to diesel. Eur. Respir. J. 2004, controlled scanning electron microscopy applied to an ambient urban 475
408 23, 82–86. aerosol sample. Aerosol Sci. Technol. 2001, 34, 97–107. 476
409 (3) Dockery, D.; Pope, C. A., III; Xu, X.; Spengler, J. D.; Ware, J. H.; (27) Kim., D.; Hopke, P. K. Classification of individual particles 477
410 Fay, M. E.; Ferris, B. J.; Speizer, F. E. An association between air based on computer-controlled scanning electron microscopy data. 478
411 pollution and mortality in 6 US cities. N. Engl. J. Med. 1993, Aerosol Sci. Technol. 1988, 9, 133–151. 479
412 329, 1753–1759. (28) Kim., D.; Hopke, P. K.; Massart., D. L.; Kaufman., L.; Cassucio., 480
413 (4) Vedal, S.; Brauer, M.; White, R.; Petkau, J. Air pollution and daily G. S. Multivariate analysis of ccsem auto emission data. Sci. Total 481
414 mortality in a city with low levels of pollution. Environ. Health Perspect. Environ. 1987, 59, 141–155. 482
415 2003, 111, 45–51. (29) Hopke, P. K.; Song, X. H. Classification of single particles by 483
416 (5) Bell, M. L.; Samet, J. M.; Dominici, F. Time-series studies of neural networks based on the computer-controlled scanning electron 484
417 particulate matter. Ann. Rev. Public Health 2004, 25, 247–280. microscopy data. Anal. Chim. Acta 1997, 348, 375–388. 485

H dx.doi.org/10.1021/es103831w |Environ. Sci. Technol. XXXX, XXX, 000–000


Environmental Science & Technology ARTICLE

486 (30) WebMet.com (cited 1/05/2010); http://www.webmet.com/


487 met_monitoring/663.html.
488 (31) Carpenter, G. A.; Grosberg, S.; Rosen, D. B. ART 2-A: An
489 adaptive resonance algorithm category learning and recognition. Neural
490 Networks 1991, 4, 493–495.
491 (32) Xie, Y.; Hopke, P. K.; Wienke, D. Airborne particle classification
492 with a combination of chemical composition and shape index utilizing an
493 adaptive resonance artificial neural network. Environ. Sci. Technol. 1994,
494 28, 1921–1928.
495 (33) Wang, Y.; Huang, K.; Li, C.; Mi, H.; Luo, J.; Tsai, P. Emissions
496 of fuel metals content from a diesel vehicle engine. Atmos. Environ. 2003,
497 37, 4637–4643.
498 (34) Councell, T. B.; Duckenfield, K. U.; Landa, E. R.; Callender, E.
499 Tire-wear particles as a source of zinc to the environment. Environ. Sci.
500 Technol. 2004, 38, 4206–4214.
501 (35) Gustafsson, M., Blomqvist, G., Dahl, A., Gudmundsson, A.,
502 Ljungman, A., Lindbom, J., Rudell, B., Swietlicki, E. Inhalable particles
503 from the interaction between tire, road surface and friction material; VTI
504 report 520, VTI, SE-581 95; Link€oping (in Swedish with English
505 abstract), 2005.
506 (36) Gustafsson, M.; Blomqvist, G.; Gudmundsson, A.; Dahl, A.;
507 Jonsson, P.; Swietlicki, E.; Bohgard, M.; Lindbom, J.; Ljungman, A.
508 Properties and toxicological effects of particles from the interaction
509 between tyres, road pavement and winter traction material. Sci. Total
510 Environ.. 2008, 393, 226–240.
511 (37) Gustafsson, M.; Blomqvist, G.; Gudmundsson, A.; Dahl, A.;
512 Jonsson, P.; Swietlicki, E. Factors influencing PM10 emissions from road
513 pavement wear. Atmos. Environ. 2009, 43, 4699–4702.
514 (38) Schauer, J. J.; Lough, G. C.; Shafer, M. M.; Christensen, W. F.;
515 Arndt, M. F.; DeMinter, J. T.; Park, J. S. Characterization of Metals
516 Emitted from Motor Vehicles; HEI Report 133; Health Effects Institute,
517 2006; http://pubs.healtheffects.org/view.php?id=150.
518 (39) Newman, E. I. Phosphorous inputs to terrestrial ecosystems.
519 J. Ecol. 1995, 83, 713–726.
520 (40) Gehrig, R.; Lienemann., H. M.; Zwicky, P.; Bukoweicki., C. N.;
521 Weingartner., N.; Baltensperger, E. U.; Buchmann, B. Contributions of
522 railway traffic to local PM10 concentration in Switzerland. Atmos.
523 Environ. 2007, 41, 923–933.
524 (41) Wongphatarakul, V.; Friedlander, S. K.; Pinto, J. P. A comparative
525 study of PM2.5 ambient aerosol chemical databases. Environ. Sci. Technol.
526 1998, 32, 3926–3934.
527 (42) Wilson, J. G.; Kingham, S.; Pearce, J.; Sturman, A. P. A review of
528 intraurban variations in particulate air pollution: Implications for
529 epidemiological research. Atmos. Environ. 2005, 39, 6444–6462.
530 (43) Henry, R. C. Receptor model applied to patterns in space
531 (RMAPS) 0.1. Model description. J. Air Waste Manage. Assoc. 1997,
532 47, 216–219.
533 (44) Paatero, P.; Hopke, P. K.; Hoppenstock, J.; Eberly, S. I.
534 Advanced factor analysis of spatial distributions of PM2.5 in the eastern
535 United States. Environ. Sci. Technol. 2003, 37, 2460–2476.

I dx.doi.org/10.1021/es103831w |Environ. Sci. Technol. XXXX, XXX, 000–000

You might also like