ARTICLE IN PRESS

1352-2310/$ - se

doi:10.1016/j.at

�CorrespondE-mail addr

Atmospheric Environment 42 (2008) 1064–1069

www.elsevier.com/locate/atmosenv

Technical note

City-wide sweeping a source for respirable particulate matterin the atmosphere

Ankit Tandon, Sudesh Yadav, Arun K. Attri�

School of Environmental Sciences, Jawaharlal Nehru University, New Delhi 110067, India

Received 4 October 2007; received in revised form 3 December 2007; accepted 3 December 2007

Abstract

Most of the cities located in Northern India are afflicted with the presence of unusually high concentration of PM10 in

the ambient environment posing a serious risk to human health. To understand the reasons underlying the persistence of

the high levels of PM10 in the Delhi region, a novel experiment was designed by appropriating a well-known tracer-

source—Diwali fireworks—emitting a large amount of particulate matter (PM) in the atmosphere. Sequential eight hourly

PM10 samples were collected and analyzed for the elemental signatures associated with the tracer and other sources.

Principal component analysis was used to resolve the sources; their respective mass contribution to PM10 load, in time

sequence, was estimated using absolute principal component score method. The results suggest that the well-established

practice of city-wide street-cleaning, resuspends the surface deposited PM10 back to the atmosphere. We suspect that this

practice resuspends about 25% of the sedimented PM10 back into the atmosphere.

r 2007 Elsevier Ltd. All rights reserved.

Keywords: PM10; APCS; Tracer source; Resuspension

1. Introduction

Respirable particulates having aerodynamic dia-meter p10mm (PM10) are an important part of theatmosphere. PM10, when present in excess of50mgm�3 are known to adversely affect humanhealth (WHO, 2006); epidemiological evidence in-dicate that even a low level of exposure leads to anincrease in the risk factors for cardiopulmonarydiseases, stressed respiratory physiology, mortalityand morbidity (Pope, 2000). Also, the physical andchemical attributes of the particulate matter (PM)have a significant influence on many atmospheric

e front matter r 2007 Elsevier Ltd. All rights reserved

mosenv.2007.12.006

ing author. Tel.: +91 11 26704309.

ess: attri@mail.jnu.ac.in (A.K. Attri).

processes: visibility, chemistry and radiative balance(Dickerson et al., 1997). Their concentration andcompositional variability in the atmosphere, in thespatial and temporal sense, are closely linked tothe nature and strength of the emitting source(s) andthe meteorological parameters. In recent years, manycities located in the northwest region of India areafflicted with the persistence of high PM10 concen-trations in the atmosphere, which has raised con-siderable concern in view of the ensued health effects.The case considered here is the national capitalregion of Delhi, which has the dubious distinction ofbeing one of the most polluted cities in the world—where a 24h average PM10 load o150mgm�3 is ararity; only rainfall assists in the lowering of theirambient concentration (CPCB, 2007). Past studies to

.

ARTICLE IN PRESSA. Tandon et al. / Atmospheric Environment 42 (2008) 1064–1069 1065

identify the anthropogenic sources responsible forthe sustained high PM10 load have identifiedvehicular and industrial activities as the maincontributors (Kumar et al., 2001; Khillare et al.,2004; Karar and Gupta, 2007). It is important tostate here that the city’s roads, streets and by-lanesare swept every morning by close to a hundredthousand personnel. Large brooms are used in a‘‘whisking’’ motion to agitate the surface and collectthe surface deposited litter in a pile. In this process,the surface-sedimented material mushrooms into theatmosphere, visible as a dusty cloud to any casualobserver. The question we asked was: ‘‘Does thispractice, spanning the city, act as a potential PM10

source by causing their resuspension back into theatmosphere?’’ To address this question, we designeda novel experiment.

It was crucial for this experiment to select aknown episodic event emitting a large amount ofPM10 in the atmosphere. Prior knowledge concern-ing the timing of this episodic source, being activeand quiescent, was required to use it as a tracer-source. These traits would help in tracking thepresence of elements emitted by the tracer-source inthe composite PM10 load. Collection of PM10

samples could then be planned around the timingof the tracer-source’s emission activity. The Diwalifestival, which is marked with fireworks, was usedas a tracer-source in our investigation for thefollowing reasons: (a) emissions from the fireworkswould carry element-specific signatures in them—e.g. metal oxidizers (Na and K), color and sparkleemitters (Al, Mg, Cu and Sr)—(Attri et al., 2001);(b) the timing of the start and finish of fireworkswould be known; (c) vehicular traffic and industrialactivities are minimal during the festival andthe day after; (d) the morning after the fireworksthe daily practice of city-wide cleaning is under-taken at a much larger scale to remove the litterfrom the previous night’s fireworks activity; and(e) the date of the festival coincides with theonset of the winter season, marked with calm winds(1.4771.13m s�1) and little variations in othermeteorological parameters (temperature, 22.371.0;RH, 63.578.2) in the region. The location selectedfor this experiment was a known receptor site(Jawaharlal Nehru University campus), wherefirework activity during Diwali is absent andvehicular traffic is low compared to the surroundingregions of the city (Singh et al., 1997; Attriet al., 2001). To the best of our knowledge thisis the first study of its kind, where by designing a

novel experiment around a tracer-source, the city-wide cleaning practice has been investigated forits role in the resuspension of PM10 back to theatmosphere.

2. Materials and methods

On Diwali and for 6 days before and after it (26October to 7 November 2005), samples of PM10

were collected at eight hourly intervals in timeseries sequence, on EPM 2000 filters using respir-able dust samplers (Envirotech model-APM460BL). Sequential samples of PM10 were collectedat 15m height; major roads were at 1.5 km distancefrom the sampling site. The aim was to use thepresence of elements in the emissions fromthe fireworks as tracer-source signature. A total of39 PM10 samples (n) for 15 elements (m) wereanalyzed (Al, Ba, Ca, Cd, Cr, Cu, Fe, K, Mg, Mn,Na, P, Pb, Sr, and Zn) by inductively coupledplasma atomic emission spectrometer (JY France,model Ultima 2). The concentration of respectiveelements in all samples (ngm�3) was used tocalculate Varimax rotated principal componentloadings to identify the number of principalcomponents (p) having eigenvalue X1.0, involvedin the emission of PM10 (Wilks, 2006). Absoluteprincipal component scores (APCSs) (details givenin supplementary material) were calculated usingthe method developed by Thurston and Spengler(1985).

3. Results and discussion

On the basis of PCA and APCS analysis(Thurston and Spengler, 1985), four differentsources (S1–S4) contributing to the PM10 load wereidentified (Table 1 and Fig. 1A–D). High compo-nent loadings associated with Al, Mg, K, Cu, Sr,and Na were identified as source 1 (S1), whichaccounted for 33.01% variance. Presence of theseelements and timing of their emission coincided withthe onset of fireworks activity, ascertaining thatthese emissions were from the fireworks (Attri et al.,2001); source 2 (S2) explained 18.44% variance andwas associated to the local crustal material (Thur-ston and Spengler, 1985); source 3 (S3) explained16.50% variance and represented crustal materialtransported through wind from adjoining miningareas (Yadav and Rajamani, 2006) and source 4(S4) reported 15.09% variance and was linked to thevehicular emissions (Kumar et al., 2001). In India,

ARTICLE IN PRESS

Table 1

Varimax rotated principal component loadings of 15 elements present in PM10 samples and their enhancement factors (EFi) in 168, 176

and 184h samples

Elements

analyzed

Varimax rotated principal

component loadings

Average reference

mass concentration

(ngm�3)

EFi, when

tracer source

was active

(168 h)

EFi, when

tracer source

was inactive

(176 h)

EFi, when tracer

source was inactive

and city-wide

cleaning was active

(184 h)

S1 S2 S3 S4

Al 0.96 0.20 0.02 0.11 6917851 58 4 15a

Mg 0.94 0.27 0.02 0.14 2087135 25 2 6a

K 0.96 0.19 0.07 0.06 358271448 24 2 8a

Cu 0.61 �0.09 0.62 0.23 2379 14 4 2

Sr 0.70 0.47 0.23 0.25 1679 8 5 3

P 0.49 0.61 0.06 0.31 48730 4 2 2

Na 0.73 0.39 0.32 �0.14 34167377 3 1 2a

Ca 0.22 0.83 0.02 0.18 10417791 2 1 1

Fe 0.28 0.70 �0.12 0.20 3437219 4 1 7a

Ba 0.09 0.72 0.56 0.03 5397360 2 2 2

Cd 0.45 0.12 0.62 0.31 1.771 10 2 3a

Zn 0.45 0.22 0.67 0.45 589775.4 3 1 1

Pb �0.12 �0.04 0.84 0.05 1307119 1 1 1

Mn 0.14 0.13 0.14 0.92 21710 5 1 2a

Cr 0.02 0.32 0.22 0.88 2.171.3 5 1 1

Eigenvalues 4.95 2.77 2.47 2.26

Percent of

variance

33.01 18.44 16.50 15.09

Source S1 was identified as tracer-source (fireworks). Source 2 was linked with emissions from crustal material. Source S3 represented wind

assisted transported contributions arising from the mining activity located 100 km away from the receptor site. Emissions from source 4

were associated with vehicular traffic.

Component loadings in bold (�0.5 and above) associate corresponding elements with the source.aElements detected in resuspension.

A. Tandon et al. / Atmospheric Environment 42 (2008) 1064–10691066

the use of Pb was discontinued as an additive ingasoline in 2001–2002 and this explains the lowloading of Pb in S4; manganese (Mn) basedadditives in gasoline are used to increase gasolineoctane, and Mn and Cr are emitted by vehicles asbrake-dust. Time-dependent emission activity ofthese sources is evident (Fig. 1A–D). PredominantPM10 emissions from S1 (548 mgm�3; 72% of PM10

load) occurred in 168 our sample (4 p.m. to 12midnight). A significant decrease in S1’s emissions(176 h sample) overlapped the span when fireworkactivity stopped (47 mgm�3; 20% of PM10 load). Allsubsequent PM10 samples were collected during theperiod when no fireworks were burnt. However,increased S1’s emissions (118 mgm�3; 41% of PM10

load) were registered again in the 184 h sample:collected between 8 a.m. and 4 p.m., the periodwhen the entire city was swept to clean post-fireworks litter (Fig. 1A). Emissions arising fromS2, local crustal material, were identified in allsamples (Fig. 1B). Source S3 was active throughout,

samples collected at 48 and 56 h registered asignificant increase in the emissions from S3 (224and 203 mgm�3, respectively) (Fig. 1C). The pre-sence of Cu, Zn and Pb in these samples was thesignatures of the wind assisted transport of dustarising from the mines located 100 km southwest ofthe receptor site; ascertained from air back-trajec-tories using NOAA HYSPLIT model (http://www.arl.noaa.gov/ready/). Emissions from the ve-hicular source (S4) remained low throughout, but asharp increase in emissions from S4 (102 mgm�3)occurred between 128 and 136 h, i.e. 2 dayspreceding the festival night (Fig. 1D). At this time,a large influx of traffic into the Delhi region occurs:vehicles throng the streets, much akin to theChristmas shopping rush in the West. Comparedto the other sources (S1–S3) the emissions from S4were significantly low (Fig. 1D).

Observed temporal variability in the total elementalload—the sum of the loads of 15 elements analyzed ineach PM10 sample—is plotted as a proportion of the

ARTICLE IN PRESS

600

500

400

300

200

100

-100

600

500

0

400

300

200

100

0

-100

(µg.

m-3

)

0 24 48 72 96 120 144 168 192 216 240 264 288 312

Hrs

0 24 48 72 96 120 144 168 192 216 240 264 288 312

Hrs

(µg.

m-3

)

Fig. 1. Absolute PM10 mass concentrations calculated by APCS method, emitted by the four identified sources (S1–S4) in each sample are

plotted (Thurston and Spengler, 1985; Wilks, 2006). Panel A shows the emission profile of tracer-source (fireworks). The maximum

emissions occurred at 168 h. The resuspension of the surface deposited PM10 load is apparent in the sample collected at 184 h. The source

S2 is associated with the local crustal material, and these emissions are present uniformly throughout the study period (Panel B). Panel C

represents the PM10 load arising from wind transported material. Contribution to the PM10 from vehicular traffic (source S4) was

observed to be minimal during this study (Panel D). However, 128 h sample shows significant contribution from S4.

A. Tandon et al. / Atmospheric Environment 42 (2008) 1064–1069 1067

total load of the respective PM10 sample (Fig. 2).Observations from three samples in sequence (168, 176and 184h) show a sharp increase in the proportion ofthe total elemental mass (149mgm�3; 19%) when thetracer-source was active, followed by a decrease(17mgm�3; 6%) and then a significant increase again(51mgm�3; 26%). It is interesting to note, here,that with in a time span of 8h, large amount oftotal elemental load (149�17 ¼ 132mgm�3) emittedduring fireworks sedimented; which accounts for 89%of the total elemental load present in 168h sample(Fig. 2).

The increased elemental mass proportion in the184h sample, when the tracer-source was inactive, wasimportant to note. This increase could only arise if thesedimented elemental mass deposits, subsequent totheir emission from the tracer-source, are somehowlifted up in the atmosphere. The resuspended totalelements load (51�17 ¼ 34mgm�3) accounted for

(34/132� 100 ¼ ) 26% of the total load, which hadsedimented during the collection of 176h sample. Thetiming of this observed increase in the elemental masscoincides with the city-wide cleaning, which supportsthe assertion that this practice maybe causing theresuspension of the surface deposits back into theatmosphere. If this be the case, then the resuspendedPM10 load will carry composite elemental signaturesspecific to the tracer-source plus other sources; akin tothe sedimented deposits. To validate this assertion, wecalculated enhancement factors (EFs) for each elementin three samples (168, 176 and 184h) (Table 1,columns 7–9). EFi for each element was calculatedby using following equation: EFi ¼ X ij=X i, where Xij

is the mass concentration (ngm�3) for ith element(i ¼ 1–15) in jth sample (j ¼ 1–39) and X i is the meanmass concentration (ngm�3) for an element present infirst three and last three samples (total six) of the timeseries. The X i represents the average reference mass

ARTICLE IN PRESS

100%

90%

80%

70%

60%

50%

40%

30%

20%

10%

0%

8 24 40 56 72 88 104

120

136

152

168

184

200

216

232

248

264

280

296

312

Hours

Elemental Mass Rest of the PM10 Mass

Fig. 2. Total mass (mgm�3) of 15 elements analyzed in each sample is shown in a proportion to the rest of the mass (mass of the PM10

sample�sum of the masses of 15 elements present in the sample). An increase in the proportion of total elemental mass is apparent in the

sample collected at 168 (4:00 p.m. to 12:00 midnight), coinciding with the activity of the tracer-source (fireworks). The next sample,

collected at 176 h (12:00 midnight to 8:00 a.m.), show a decline due to the sedimentation of PM10 load from the atmosphere to the ground

surface. Subsequent increase in the proportion of total elemental mass observed in 184 h sample (8:00 a.m. to 4:00 p.m.) coincided with the

city-wide sweeping to remove litter from the fireworks. The proportion of total elemental mass in this sample increased substantially.

A. Tandon et al. / Atmospheric Environment 42 (2008) 1064–10691068

concentration (ngm�3), assumption is that 5 daysbefore and after the firework’s event there will be noactivity of tracer source (S1) and emission strengths ofother sources (S2–S4) will be as of a normal winterday. The EFs of elements show (Table 1, columns7–9): (i) tracer-source’s activity, (ii) sedimentation oftracer-source specific elements and (iii) resuspendedmaterial bearing signatures of all the identified sources(S1–S4).

4. Conclusions

We conclude that any contribution to PM10 loadarising from the practice of sweeping will dependupon the amount as well as the nature of PM10

deposited previously on the street surfaces, and it isevident from the extent of variability in EFs ofelements specific to the tracer-source and othersources (Table 1, column 9). The resuspended PM10

will bear composite signatures of all other sources ifthe experiment was done in conventional manner.Thus, the detection of street cleaning as a sourcewould be difficult if the experiment did not involvethe use of a known tracer source. In view of theresults from our investigation, we suspect that thisdaily practice of city-wide sweeping will assist in thepersistence of PM10 load over a longer time span by

resuspending the surface deposits. As this practice iswidely prevalent in most of the cities located inIndian subcontinent, it is likely that the resuspen-sion of PM10 in the ambient environment assists inthe persistence of high PM10 load. Also, thedeveloped strategy of using a known tracer-sourceprovides a novel approach to investigate the role ofother anthropogenic activities, so far consideredbenign, to air pollution.

Acknowledgments

Authors thank Prof. V. Rajamani, School ofEnvironmental Sciences, JNU, New Delhi, forproviding ICP-AES facility for elemental analysis.Financial assistance provided by Council of Scien-tific and industrial Research (CSIR) India, in theform of a research project, is acknowledged.Authors acknowledge the two anonymous refereesfor their appraisal of the manuscript and valuablecomments.

Appendix A. Supplementary material

Supplementary data associated with this articlecan be found in the online version at doi:10.1016/j.atmosenv.2007.12.006.

ARTICLE IN PRESSA. Tandon et al. / Atmospheric Environment 42 (2008) 1064–1069 1069

References

Attri, A.K., Kumar, U., Jain, V.K., 2001. Formation of ozone by

fireworks. Nature 411, 1015.

Central Pollution Control Board (CPCB, 2007), Air Quality of

Delhi. Available at/http://www.cpcb.nic.in/bulletin/bul.htm/S.Dickerson, R.R., et al., 1997. The impact of aerosols on solar

ultraviolet radiation and photochemical smog. Science 278,

827–830.

Karar, K., Gupta, A.K., 2007. Source apportionment of PM10 at

residential and industrial sites of an urban region of Kolkata,

India. Atmospheric Research 84, 30–41.

Khillare, P.S., Balachandran, S., Bharat, R.M., 2004. Spatial and

temporal variation of heavy metals in atmospheric aerosol of

Delhi. Environmental Monitoring and Assessment 90, 1–21.

Kumar, A.V., Patil, R.S., Nambi, K.S.V., 2001. Source apportion-

ment of suspended particulate matter at two traffic junctions in

Mumbai, India. Atmospheric Environment 35, 4245–4251.

National Oceanic Atmospheric and Administration (NOAA),

NOAA HYSPLIT MODEL. Available at /http://www.arl.noaa.gov/ready/S.

Pope III, C.A., 2000. Review: epidemiological basis for particu-

late air pollution health standards. Aerosol Science and

Technology 32, 4–14.

Singh, A., Sarin, S.M., Shanmugam, P., Sharma, N., Attri, A.K.,

Jain, V.K., 1997. Ozone distribution in the urban environ-

ment of Delhi during winter months. Atmospheric Environ-

ment 31, 3421–3427.

Thurston, G.D., Spengler, J.D., 1985. A quantitative assessment

of Source contributions to inhalable particulate pollu-

tion in metropolitan Boston. Atmospheric Environment

19, 9–25.

Wilks, D.S., 2006. Statistical Methods in the Atmospheric

Sciences. Academic Press, San Diego, pp. 463–508.

WHO (2006), Use of air quality guidelines in protecting public

health: global update. Available at /http://www.who.int/

mediacentre/factsheets/fs313/enS.

Yadav, S., Rajamani, V., 2006. Air quality and trace metal

chemistry of different size fractions of aerosols in N-NW

India—implications for the source diversity. Atmospheric

Environment 40, 698–712.