Quantifying the Impact ofResidential Heating on the UrbanAir Quality in a Typical EuropeanCoal Combustion RegionH E I K K I J U N N I N E N , † , ‡ J A C O B M Ø N S T E R , †

M A R I A R E Y , † J O S E C A N C E L I N H A , †

K E V I N D O U G L A S , † M A T T H E W D U A N E , †

V I C T T O R I O F O R C I N A , † A N N E M U L L E R , †

F R I T Z L A G L E R , † L U I S A M A R E L L I , †

A N N E T T E B O R O W I A K , †

J O A N N A N I E D Z I A L E K , † , §

B O S T I A N P A R A D I Z , †

D A N I E L M I R A - S A L A M A , † J O S E J I M E N E Z , †

U T E H A N S E N , † C O V A D O N G A A S T O R G A , †

K R Z Y S Z T O F S T A N C Z Y K , | M A R V I A N A , ⊥

X A V I E R Q U E R O L , ⊥

R A C H E L L E M . D U V A L L , #

G A R Y A . N O R R I S , # S T E F A N T S A K O V S K I , ∇

P E T E R W Å H L I N , O J I R I H O R A K , † , [ A N DB O R . L A R S E N * , †

European Commission, Joint Research Centre, 1IES/10IHCP,1Transport and Air Quality/10Chemical Assessment andTesting, Via Enrico Fermi, 2749 - Ispra (VA), 21020 - Italy,University of Helsinki, P.O. Box 64, Helsinki 00014, Finland,Malopolski Voivodship Inspectorate for Environ. Protection,pl. Szczepanski 5 - Krakow, 31-011- Poland, Central MiningInstitute, Plac Gwarkow 1, - 40-166 Katowice, Poland,Institute of Earth Sciences, ‘Jaume Almera’, CSIC, C/ Lluis Solei Sabarıs S/N 08028 Barcelona, Spain, U.S. EnvironmentalProtection Agency, Office of Research and Development,National Exposure Research Laboratory, Research TrianglePark, North Carolina 27711, Faculty of Chemistry, Universityof Sofia, J. Bourchier Blvd. 1, 1164 Sofia, Bulgaria, NationalEnvironmental Research Institute, Department of AtmosphericEnvironment, Frederiksborgvej 399, P.O.Box 358 - Roskilde,4000 Denmark, and VSB -Technical University of Ostrava,Energy Research Center, 17 listopadu 15/2172, 708 33Ostrava-Poruba, Czech Republic

Received January 12, 2009. Revised manuscript receivedJuly 21, 2009. Accepted September 1, 2009.

The present investigation, carried out as a case study in atypical major city situated in a European coal combustion region(Krakow, Poland), aims at quantifying the impact on theurban air quality of residential heating by coal combustion incomparison with other potential pollution sources such as powerplants, industry, and traffic. Emissions were measured for 20major sources, including small stoves and boilers, and the

particulate matter (PM) was analyzed for 52 individualcompounds together with outdoor and indoor PM10 collectedduring typical winter pollution episodes. The data were analyzedusing chemical mass balance modeling (CMB) and constrainedpositive matrix factorization (CMF) yielding source apportion-ments for PM10, B(a)P, and other regulated air pollutants namelyCd, Ni, As, and Pb. The results are potentially very useful forplanning abatement strategies in all areas of the world, wherecoal combustion in small appliances is significant. Duringthe studied pollution episodes in Krakow, European air qualitylimits were exceeded with up to a factor 8 for PM10 and upto a factor 200 for B(a)P. The levels of these air pollutants wereaccompanied by high concentrations of azaarenes, knownmarkers for inefficient coal combustion. The major culprit forthe extreme pollution levels was demonstrated to be residentialheating by coal combustion in small stoves and boilers(>50% for PM10 and >90% B(a)P), whereas road transport(<10% for PM10 and <3% for B(a)P), and industry (4-15% forPM10 and <6% for B(a)P) played a lesser role. The indoorPM10 and B(a)P concentrations were at high levels similar tothose of outdoor concentrations and were found to have thesame sources as outdoors. The inorganic secondary aerosolcomponent of PM10 amounted to around 30%, which for a largepart may be attributed to the industrial emission of theprecursors SO2 and NOx.

IntroductionThe present investigation aims at quantifying the impact ofresidential heating by coal combustion on urban air qualityin comparison with other potential pollution sources suchas power plants, industry and traffic in a major cityrepresentative of European coal combustion regions. Withits number of activities using coal as fuel, the area of Krakow(Poland) has emission sources typical for many areas in thenew EU member states and major countries like China andIndia. Thus, information acquired in Krakow is not only usefulfor the planning of future abatement strategies for thismetropolitan area, but may also be valuable for the designof pollution control and policy strategies in similar metro-politan areas in Europe and Asia.

A number of studies have tried to estimate the impact ofcoal combustion on smog episodes in Polish and othereastern European cities (1-5). However, none of these studiesincluded chemical analysis of the PM emitted from sourcesand therefore they have not succeeded in distinguishingamong the various coal combustion sources in the domestic,industrial, and transport sectors. The present investigationwas planned to address this issue utilizing receptor modelingtechniques for source apportionment. During two typicalwinter episodes with stable meteorological conditions, lowwinds, and shallow inversion layers, PM10 was collected fromsampling sites representing rural, urban, industrial, and trafficrelated air quality conditions. In addition, PM emitted bycombustion processes was sampled from 20 representativepollution sources. The collected samples were analyzedchemically for 52 individual source tracer compoundsincluding organics (6) and the obtained chemical fingerprintswere subjected to multivariate statistical analysis utilizingchemical mass balance modeling (CMB) and constrainedpositive matrix factorization (CMF).

The present paper presents the source contributionestimates (SCEs) for PM10 and associated toxic air pollutantsregulated in the EU, namely B(a)P, Pb, Ni, Cd, and As.

* Corresponding author phone:+39-0332-789647; e-mail: bo.larsen@jrc.ec.europa.eu.

† European Commission, Joint Research Centre.‡ University of Helsinki.§ Malopolski Voivodship Inspectorate for Environ. Protection.| Central Mining Institute.⊥ Institute of Earth Sciences.# U.S. Environmental Protection Agency.∇ University of Sofia.O National Environmental Research Institute.[ VSB -Technical University of Ostrava.

Environ. Sci. Technol. 2009, 43, 7964–7970

7964 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 20, 2009 10.1021/es8032082 CCC: $40.75 2009 American Chemical SocietyPublished on Web 09/18/2009

Experimental ProceduresStudy Area Characteristics. Krakow (50°04′ N 19°56′ E, Figure1) has more than 800,000 residents and is Poland’s secondlargest city. The city’s area of more than 300 km2 spreads onboth banks of the Vistula (Wisla) river, about 220 m abovesea level on the Malopolska Uplands at the foot of theCarpathian Mountains. Zakopane (49°18′ N 19°57′ E) lies inan expansive valley between the Tatra Mountains andGubałowka Hill approximately 100 km south of Krakow.Zakopane was selected for comparison due to its predomi-nant use of inefficient individual heating systems similar toKrakow, its remote geographical location, and its low numberof inhabitants (28,000) which rule out road transport andlocal industry as significant pollution sources. Four receptorsites were selected in Krakow and one in Zakopane. The sitesin Krakow (Figure 1) included a rural/semirural samplingstation placed in the northwest outskirts of the city (AGRI),an urban traffic station in the center of the city close to themain road (TRAFFI), an urban station placed in a district ofthe city characterized by a high number of old apartmentsheated by coal combustion in small stoves (POLI), and anindustrial site (INDU) located between the Huta IM Send-zimia Steel-works (squares 20-31) and the power plant(squares 4-6).

Source Characteristics. The major representative sourcesfor PM in Krakow, characterized for the purpose of this study,are indicated in Figure 1. The emission rates for each source,operated as close as possible to typical conditions, are listedin Table 1 of the Supporting Information (SI) for PM and airpollutants regulated by EU air quality directives, namelyB(a)P, Pb, Ni, Cd, As, SO2, CO, and NOx. The sources with thehighest PM emission rates (1-11 kg/h) are situated in thenortheastern part of Krakow, i.e., the Huta IM Sendzimiasteel-works (the iron ore sinter plant, the blast furnace, thepower plant with coke-gas and coal combustion, the pro-duction plant for fireproof material for steel production withnatural gas combustion, and the basic oxygen furnace steelplant with coke combustion), the Cementownia cementfactory (coal-fired cement kiln), and the EC-Krakow coal-

fired power plant (>50 MW). The PM emission rates fromcoal combustion in small (<5 MW) commercial boilers fittedwith rudimentary PM removal systems (e.g., a cyclone) andthe emission rates from small (<5 MW) boilers fueled withheavy oil are, as expected, much lower (150-300 g/h).However, in Krakow there are numerous small coal-firedlow-efficiency boilers (LE-boilers) distributed over the city,most of which are not fitted with any kind of abatementtechnology, which potentially makes coal-fired LE-boilersan important PM source in Krakow (7). The emission ratefrom coal- and wood-fired small stoves are the lowest inTable 1, SI (27 and 13 g/h, respectively). Nonetheless, whenthe number of flats in Krakow (approximately 20,000) heatedwith individual coal-fired small stoves (7) is taken intoconsideration this source can be of major significance.

PM Sampling. A total of 178 ambient air samples, 54source samples, and 22 indoor samples were collected. Forthe chemical analysis of organic and inorganic compoundstwo kinds of filter material were used in parallel: quartz (47mm AQFA Millipore) and mixed nitrocellulose esters (47 mm,1.2 µm pore size, Millipore). For ambient air measurements,PM10 was sampled by means of low-volume samplers (24 h,2.3 m3/h) and determined gravimetrically (8). The filters werestored at-20 °C until chemical analysis. For emission sourcemeasurements, PM was sampled iso-kinetically from theemission stacks (ISO-9096). Flue gas sampling was performedby the filter-cooler method (approx 120 °C) and conductedwith an automatic, adjusting iso-kinetic sampling system.Source sampling and characterization were carried out induplicate or triplicate at consecutive days during normaloperation conditions to ensure representative samples. Thesampling of PM10 outdoor and indoor was conducted overa 3-week period from January 16 to February 8, 2005,interrupted for 6 days (Jan. 24-28) due to heavy snowfall. Inaddition, samples were taken in four apartments (indoor aswell as outdoor on the balconies) selected close to the ambientair measurement stations (Figure 1).

Chemical Analysis. All aspects of the chemical analysisof the PM samples including quality control are described

FIGURE 1. Location of the outdoor (green circles) and indoor (blue circles) PM10 sampling sites and the characterized emissionsources (orange squares) in the Krakow metropolitan area.

VOL. 43, NO. 20, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 7965

in the SI. In brief, PAHs and azaarenes were analyzed by gaschromatography-mass spectrometry, trace elements wereanalyzed by proton induced X-ray emission, ionic specieswere analyzed by ion chromatography and atomic absorptionspectrophotometry, and elemental and organic carbon wereanalyzed by thermal-optical analysis. Furthermore, individualparticles were analyzed by the SPASS single particle massspectrometer as described in detail by Mira-Salama et al. (9).To support indoor air source apportionments, individualparticles were sampled indoor and outdoor (on the balcony)from an apartment in the city district (Figure 1) dominatedby residential heating with coal combustion in small stovesand several thousand individual particles were analyzed, eachparticle producing a positive and a negative mass spectrum.

Receptor Modeling. The resulting Krakow data set forreceptor modeling is available as SI and consists of an 88 ×52 matrix for the receptor data (88 samples with the massfraction for the 52 PM species) and a 13 × 52 matrix for thesource data (13 sources with the mass fraction for the 52 PMspecies). The source apportionment results were obtainedwith receptor models described in detail in the SI, namelyChemical Mass Balance (U.S. EPA, CMB 8.2) and ConstrainedMatrix Factorization, CMF (a hybrid of CMB and positivematrix factorization, PMF (10) especially optimized forpresent study) and intercompared to factorization modelsas described in detail elsewhere (8). Due to the relativelysmall number of samples collected during the measurementcampaign at the individual sites, the intercomparison wasfacilitated by running the models with a data set consistingof the samples from all receptor sites pooled together, which

is based upon the assumption that all receptor sites areinfluenced by the same sources. The obvious advantage ofthis approach is the gain in the number of degrees of freedom,which increases the number of factors that can be extractedwith a statistical significance. Species measured at the sametime at different sites may also yield higher correlations thanwhat would be expected if all sites were modeled individually.Although the factorization models produced results that werequalitatively in agreement with the results of CMB andCMFswhich lends credibility to the source apportionmentexercisesthe resolution that could be obtained with thelimited data set (88 samples) was not high enough for aquantitative comparison (SI).

Results and DiscussionLevels of PM and Associated Priority Air Pollutants. The24-h mean concentrations of PM10 and associated prioritypollutants during the measurement campaigns are shownin Figure 2. Detailed data are presented in the SI, whichshows that these very high PM10 concentration levelsrepresent the typical winter situation in Krakow and areconnected with meteorological episodes of low wind speedand temperature inversion. The concentrations of B(a)Pmeasured at all five stations (mean value over the twomeasurement campaigns) were high compared to the 1 ng/m3 EU target value (yearly average): POLI: 56 ng/m3, TRAFFI:20 ng/m3, INDU: 24 ng/m3, AGRI: 29 ng/m3, ZAKO: 56 ng/m3

(maximum value)200 ng/m3). Only a few pollution episodesper year similar to the ones encountered in the present studyare enough to cause infringement of the 1 ng/m3 yearly

FIGURE 2. Concentration levels of PM10, B(a)P, Pb, Cd, Ni, and As at the five measurement sites during the two campaign weeks.

7966 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 20, 2009

average target value. The sum of all measured PAHs reachedlevels of up to 2000 ng/m3 (400 ng/m3 expressed as B(a)Ptoxicity equivalents) and the sum of the measured azaarenesin the five measurement sites on average for the twocampaigns were: POLI 34 ng/m3; TRAFFI 18 ng/m3; INDU 24ng/m3; AGRI: 19 ng/m3; ZAKO: 60 ng/m3 (maximum value)128 ng/m3). These levels are 1-2 orders of magnitude higherthan the highest levels previously reported for polluted areasin Europe and the United States (11-14). Like other PACs,some azaarenes are known mutagens and/or carcinogensand although there are no air quality limits for azaarenes,the present findings may be a cause for concern.

The concentration levels of lead, arsenic, and nickelmeasured at all five stations (Figure 2) do not pose problemsfor the EU limits (15).

The average concentrations of PAC sorbed on PM (µgPAC/g PM10) during the two campaigns were very similar forall sites indicating that one source type is dominating. Thesubstantial concentrations of azaarenes points to coalcombustion (14) being this source type. The relative con-tribution of azaarenes was highest in Zakopane, in particularfor dibenzo[a,h]acridine, which is to be expected if homeheating by inefficient combustion of coal is the main sourcein this area. In metropolitan areas, air pollution scientiststypically find the highest PAC concentrations in PM nearmain roads and certain five-six ring PAHs can even be usedin source apportionment studies as organic markers forgasoline-fueled-vehicle exhaust (16, 17). In the present studythe PM from the TRAFFI site did not contain the highest PACconcentrations, which makes it evident that in areas domi-nated by coal combustion such markers are not very usefulfor earmarking traffic, and a multicomponent receptormodeling approach is called for.

Source Contribution Estimates. The source apportion-ment involved procedural choices for the selection of sourcecomposition profiles and fitting species in the CMB and CMFcalculations, which are all discussed in detail in the SI.

CMB. For CMB the following source fingerprints con-tributed significantly to the observed data: residential stoves/boilers (coal), residential stoves/boilers (wood), low-efficiencyboilers (coal), low-efficiency boilers (fuel), steelworksPP(power plant), steelworks, vehicles (including exhaust as wellas brake- and tire-wear), rock, lime, cement, and road salt(NaCl). In addition to these primary sources three fingerprintswere included to account for secondary aerosol components,

namely ammonium sulfate, ammonium nitrate, and am-monium chloride.

The SCEs are shown in Figure 3 for the five sites. A goodmass apportionment was obtained for the apportioned PM10

mass levels (predicted PM10 ) 0.82 × measured PM10 + 0.3;R2 ) 0.93). The minor underestimation of the predicted PM10

may derive from intrinsic errors in the CMB approach(including analytic errors and unfulfilled basic assumptions)or from a failure to consider minor sources. It is evident inFigure 3 that on all sites the contribution to the atmosphericPM10 pollution is dominated by coal combustion in stovesand small boilers. It is also evident that the contributionfrom secondary aerosol components is much higher at theKrakow sites than in Zakopane, which strongly point to localsources, such as industry, power generation, and possiblyroad transport, which are all major emitters of SO2, NOx, orboth. As expected, sources related to road transport (trafficand resuspension) contributed most at the two sites situatednear the city center (TRAFFI > POLI) and are insignificant atother peripheral sites (AGRI and INDU) and the remote site(ZAKO). Moreover, in good agreement with what would beexpected, the sources from steelwork activities are onlyimportant at the INDU site, and home heating by combustionof wood is only important at the ZAKO site where wood iseasily available from the surrounding forested area.

Details about the temporal variation of the sourceestimates for each site and the influence of meteorologicalconditions are presented in the SI. Interestingly, CMBcomputations revealed regional transport by minor, yetsignificant, contributions from nonlocal sources to the sitesin Krakow. Hence, only during the days with westerly windssmall contributions from an industrial source (quantified as“steelworks”) were observed at the AGRI site, which is ingood accordance with the presence of an industrial area westof Krakow. Likewise, at the INDU site, the contribution fromthe steelworks activities could not be detected during thedays with winds from south, whereas the contributions fromhome heating were strongest during the meteorologicalconditions when this site was downwind of the city center.Finally, the contribution from coal combustion in smallstoves, a source type concentrated in and around the citycenter, relative to the contribution from coal combustion inLE-boilers (which is distributed all over the Krakow area)could also be explained by the meteorological conditions.Expectedly, on days with very stable conditions and insig-

FIGURE 3. Source contribution estimates (SCE) by CMB for the five sites (error bar depicts the propagated uncertainty for all SCE).

VOL. 43, NO. 20, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 7967

nificant local-range transport the contribution from smallstoves dominated at two sites near the city center (POLI andTRAFFI) and on windy days the contribution from boilersincreased. At the remote site in Zakopane these transportphenomena were not observed due to the absence of sourcesin the vicinity.

Very high concentrations of PACs were also found in theindoor air in the four investigated apartments following thesame temporal trend as the outdoor concentrations. B(a)Pconstantly exceeded the outdoor air quality limit with morethan a factor 50 during peak episodes, which speaks of thegravity of the pollution problem in this coal combustion area.The single particle analysis of outdoor and indoor air in anapartment near the POLI station revealed a striking similarityin the indoor and outdoor aerosol (SI). Both indoor andoutdoor air masses contained identical individual particlesand gave major signals for carbon with simultaneous absenceof sulfate, chlorine, and calcium, which is typical for singleparticle mass spectrometric data for coal combustion (9).The CMB results for the indoor air in the four investigatedapartments in Krakow yielded the same main sources as forthe outdoor air (from nearest receptor site) with an expectedlyhigher contribution from residential coal combustion. Similarsource implications for PAHs in indoor and outdoor air havebeen published by Naumova et al. (18). For the fourapartments the average ((SD) SCEs (µg/m3) were: residentialcoal combustion in small boilers and stoves (50 ( 20.4);secondary aerosol (5.3 ( 3.1); traffic (2.6 ( 1.9); andresuspension (1.0(0.6). The mass apportionment was higherthan 85% for all individual apartments.

CMF. A large number of exploratory model runs wereconducted with completely and partially constrained factorsfor which information was available. The selection criteriawere the optimization of mass apportionment and mini-mization of residuals for PM10 as well as single components.The most satisfactory model contained a total of 12 factors,six factors for which all elements were constrained, two factorswith some of the elements constrained, and four noncon-strained factors (SI). These factors compared well to themeasured source profiles used in the CMB model with theminor exception of the secondary aerosol components (SO4

2-,Cl-, NO3

-, and NH4+), which were not handled well by the

CMF approach with the relatively little data available,although the overall model performance for all these elementsis very good (R2 > 0.95; SI).

It was assumed that secondary aerosol contributed to allreceptor samples. Thus five common secondary aerosolcomponents were included (NH4NO3, (NH4)2SO4, NH4HSO4,NH4Cl, and H2SO4) as constrained factors by freezingconcentrations of the intrinsic compounds (NH4

+, NO3-,

SO42-, Cl-) to their respective stoichiometric ratios. Test runs

revealed better results in terms of mass coverage and r2 forpredictions by allowing the composition of these factors tovary between (2% soft-locking (see SI for explanation)intervals for all intrinsic compounds.

The large contributions from coal-combustion relatedsources made it necessary to partially constrain the profilefor vehicle emissions. This was done with the same literaturedata as in CMB, with the exception that for CMF an averagecomposition of all available profiles was used with 2 timesthe standard deviation of the averaged profiles as soft-lockinginterval in the constraints.

Road salt is a minor source and best results were obtainedby constraining this profile to the composition of sea salt.Since it is not clear how similar the road salt is to pure seasalt, large soft-locking intervals were allowed for the elementsin this constrained profile (50%-200%). In practice, withthis kind of constraint the mass ratios of the compoundsthat are known to be present in sea salt were allowed to varyin the iterations, but other compounds were blocked fromentering into the profile. The soft-locking procedure resultedin the enrichment of the profile with SO4

2-, Br-, Ca2+, Mg+,and K+, which may not only be due to a different compositionof the utilized road salt, but also may derive from road dust.

Although constraining a factorization model largelyreduces the rotational ambiguity it will not remove it totally.Remaining factors can still have rotational ambiguity amongthemselves if they have a high degree of collinearity, whichis very much the case of the Krakow data. The remainingmajor sources are likely to be combustion sources and thus,vary only little in their composition. The best results wereobtained with four nonconstrained factors together with apartly constrained profile for residential coal combustionbased on the source profile N10 (small stoves; Table 1, SI).The constrained components were EC (hard-locked, see SIfor explanation) as well as OC and PAHs ((40% soft-locked)(OC and PAHs). This containment approach yielded a factorprofile very similar to the original source profile, althoughsomewhat enriched for ammonium nitrate and soil minerals(Si, Fe, Al, and Zn) and depleted for Cl- and to some extentthe 5-6 ring PAHs. The profiles that CMF yielded for the fournonconstrained factors had Euclidian distances closest totwo CMB source profiles for coal combustion: low-efficiencyboilers (coal), steelworksPP, and two CMB profiles for woodcombustion: residential wood combustion in small stove (N5and N6).

The SCEs obtained by CMF are compared to those ofCMB in Table 1. Generally, good mass apportionment andpredictability (R2) are observed for both models with agenerally good agreement in the estimated SCEs. Both modelscompute the highest primary contributions to the PM10

pollution in Krakow and in particular Zakopane from home

TABLE 1. Source Contribution Estimates (±95% Confidence Interval) for PM10 in Krakow and Zakopane (Units µg/m3)

Krakow Zakopane

CMB CMF CMB CMF

home heating residential coal combustion in smallstoves and boilers

38 ( 11a 11 ( 5 43 ( 40 16 ( 16

residential heating (wood, coke, oil) 13 ( 6 46 ( 20 58 ( 31industrial power generation

(coal)LE-boilers (coal) 16 ( 3 17 ( 3 5.4 ( 3.5 5.5 ( 4.4HE-coal combustion 3.5 ( 1.2 13 ( 5 not significant 1.1 ( 0.9

secondary aerosol(inorganics)

sulfates, nitrates and chlorides 16 ( 2 16 ( 2 7.7 ( 3.7 9.4 ( 4.4

traffic and resuspension vehicles 5.8 ( 2.0 3.7 ( 1.5 not significant 0.5 ( 0.4resuspension (incl. road salt) 2.1 ( 0.3 2.0 ( 0.3 not significant 1.2 ( 0.4mass coverage 82% 84% 91% 82%R2 0.94 0.96 0.89 0.89

a In a large number of CMB runs, profiles for residential heating (coal, wood, coke, oil) resulted collinear and wereestimated as coal.

7968 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 20, 2009

heating with some differences in the breakdown of the SCEsinto combustion of coal, and wood/coke/oil due to thediscussed collinearity of these sources. In Krakow the SCEsfrom this source category (mainly coal combustion) cor-responded to 30-50% and in Zakopane (mainly wood andcoal) to 80-90%, which is in accordance with high numberof small stoves in Krakow and Zakopane. The second highestprimary contribution was estimated by both models to comefrom industrial power generation (coal), with SCEs of 30-40%in Krakow and 5-10% in Zakopane (mainly wood and coal).Within this category combustion of coal in low-efficiencyboilers was the major source. When the PM emission ratesfrom the HE-coal combustion sources are taken intoconsideration (Table 1, SI) this finding may seem surprising.However, the HE-coal combustion sources are all emittingthrough very high stacks, which are constructed to ensureminimum fallouts in the Krakow area. In addition, duringthe measurement campaigns the mixed boundary layer (MBL)was often so shallow that the stack emissions occurred abovethe MBL (8). Traffic and resuspension was estimated by bothmodels as a minor primary source with SCEs of 8-10% inKrakow and less than 2% in Zakopane. At first thought thisfinding may appear surprising for a metropolitan area.However, it should be seen in the light of large intensities ofthe coal combustion sources in the area. The absolute valueof SCE in the range of 3.7-5.8 µg/m3 obtained for Krakowis in the same order of magnitude as road traffic SCEs inmany European metropolitan areas such as, e.g., Zurich(19, 20). The contribution from secondary aerosols wasestimated by both models to be 20-21% in Krakow and lessthan 8-10% in Zakopane. Secondary aerosols are formed inthe air by chemical transformations of gaseous pollutantsduring transport to the receptor site, and as such are muchbetter dealt with by source-oriented chemical transportmodels (21, 22). Receptor modeling cannot attribute sourcesfor the proportion of PM made up of secondary aerosol, andthe source factors which contain high loadings of SO4

2-, NO3-,

and NH4+ may be interpreted as secondary aerosol. The SCE

of 16 µg/m3 inorganic secondary aerosol in Krakow is highcompared to most other European data (22, 23), and stronglyindicates that a significant proportion of this source is local/regional rather than remote. This may derive from all sources,which the applied receptor models cannot apportion.However, based on the very high emission factors for SO2

and NOx measured in the present study for industrial powergeneration (Table 1, SI) it is likely that this source categoryis a major contributor to secondary aerosols. Only theinorganic part of secondary aerosol could be quantified bythe CMB and the CMF approach. It is well-known that organiccarbon emitted in the gas-phase at relatively high sourcetemperatures may condense onto existing PM in the atmo-sphere, and thus, a part of the PM mass attributed to primarysources in the present study may actually derive as secondaryaerosol. The use of aerosol mass spectrometry as demon-strated by the group of Prevot (21, 24) may be useful inresolving this ambiguity in the OC apportionment.

Receptor models can produce SCEs not only for the PM10

mass, but also for the individual chemical compounds.

Whereas the SCEs for the PM10 mass are obtained in aniterative process, which aims to minimize the overall dif-ference (sum of least-squares) between measured andmodeled concentrations of all the included receptor com-pounds, this does not imply that the solutions are optimalfor each individual compound. Thus the value of sourceapportionment from single compounds depends very muchon the model performance for these compounds. Notsurprisingly, CMF provided the best performance for indi-vidual species as this is part of its optimization process (CMBoptimizes percentage mass, goodness of fit, and fractionestimate as well). Nevertheless it remains an open questionwhether this is an artifact of the method and whether theCMF single species fitting procedure actually yields moreaccurate SCEs for a particular individual species than CMB.For B(a)P the single compound CMF source apportionmentyielded very good agreements between observed and mod-eled concentrations (R ) 0.99; SI) and revealed, that for allreceptor sites, residential heating is the dominant source,which together with low-efficiency boilers contribute withmore than 90% for this toxic air pollutant. The remaining10% derives mainly from high-efficiency coal combustion(e.g., power generators). The contribution from road transport(+ resuspension) is only significant at the traffic site, andeven here it contributes less than 3% (Figure SI-11). Incomparison, source apportionment studies based on PAHsource profiles in the urban and adjacent coastal atmosphereof Chicago/Lake Michigan in 1994-1995 demonstrated thatcoal combustion accounted for 48 ( 5% of the ΣPAHconcentration in both the urban and adjacent coastalatmosphere, natural gas combustion accounted for 26( 2%,coke ovens accounted for 14 ( 3%, and vehicle emissions(gas + diesel) accounted for 9 ( 4% (25).

CMF performed well for Pb (R2 ) 0.92) and Cd (R2 ) 0.92),but not for Ni (R2 ) 0.67) and As (R2 ) 0.32), which mayindicate that all sources for the latter two trace compoundsare not accounted for by the CMF model. The sourcecontributions for Pb, Cd, Ni, and As are compared in Table2 for the four main source categories as an average of allfive sites. It is seen that these compounds derive mainly fromthe industrial sources boilers and high-efficiency coalcombustion (e.g., power generation) and, although none ofthe above-mentioned heavy metals at present are found atcritical levels compared to the EU air quality limits, a futureshift in energy strategy for home heating from low-efficiencycoal combustion in small stoves and boilers to powergenerators needs careful monitoring for these heavy metals.

AcknowledgmentsWe are grateful to C. Gruning, N. R. Jensen, and P. Cavallifor single particle mass spectrometric analysis. The politicalsupport of the Polish air quality authorities as well as thelogistical support of MVIE is gratefully acknowledged.

Supporting Information AvailableMore detailed descriptions of chemical analysis, qualitycontrol, receptor models, meteorology and levels of PM and

TABLE 2. Source Apportionment for Regulated Air Pollutants in PM10: Average SCE (ng/m3) of All Five Receptor Sites (±95%Confidence Interval)

B(a)P Pb Cd Ni As

home heating (coal) 28 ( 4 17 ( 2 0.3 ( 0.04 0.2 ( 0.03 0.20 ( 0.03LE-boilers (coal) 3.4 ( 0.4 43 ( 5 1.3 ( 0.1 0.8 ( 0.1 0.9 ( 0.1HE-coal combustion 1.9 ( 0.4 25 ( 5 0.8 ( 0.2 1.1 ( 0.2 not significanttraffic and resuspension 0.04 ( 0.01 5.6 ( 1.0 not significant 0.3 ( 0.1 not significantmass coverage 99% 97% 100% 96% 85%R2 0.98 0.92 0.92 0.67 0.32

VOL. 43, NO. 20, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 7969

associated priority air pollutants. This material is availablefree of charge via the Internet at http://pubs.acs.org.

Literature Cited(1) Pastuszka, J. S. Study of PM-10 and PM-2.5 concentrations in

Southern Poland. J. Aerosol Sci. 1997, 28, 227–228.(2) Lofstedt, R. E. Transboundary environmental problems: the case

of the burning of coal in Poland for heating and electricitypurposes. Global Environ. Change 1998, 8, 329–340.

(3) Houthuijs, D.; Breugelmans, O.; Hoek, G.; Vaskovi, E.; Miha-likova, E.; Pastuszka, J. S.; et al. PM10 and PM2.5 concentrationsin Central and Eastern Europe: results from the Cesar study.Atmos. Environ. 2001, 35, 2757–2771.

(4) Kopcewicz, B.; Kopcewicz, M. Long-term measurements of iron-containing aerosols by Mossbauer spectroscopy in Poland.Atmos. Environ. 2001, 35, 3739–47.

(5) Bem, H.; Gallorini, M.; Rizzio, E.; Krzeminska, M. Comparativestudies on the concentrations of some elements in the urbanair PM in Lodz City of Poland and in Milan, Italy. Environ. Int.2003, 29, 423–428.

(6) Schauer, J. J.; Rogge, W. F.; Hildemann, L. M.; Mazurek, M. A.;Cass, G. R. T.; Simoneit, B. R. Source apportionment of airborneparticulate matter using organic compounds as tracers. Atmos.Environ. 1996, 30, 3837–3855.

(7) Jimenez, J., Niedzialek, J. , Eds. Outcome of the Krakow IntegratedProject, PM: From Emissions to Health Effects; EUR 22293 EN;European Communities: Brussels, 2006.

(8) Larsen, B. R.; Juninnen, H.; Mønster J.; Viana, M.; Tsakovski, S.;Duvall, J. R. Norris, G., Querol, X. P. The Krakow ReceptorModeling Inter-Comparison Exercise; EUR 23621 EN; EuropeanCommunities: Brussels, 2008.

(9) Mira-Salama, D.; Gruning, C.; Jensen, N. R.; Cavalli, P.; Putaud,J. P.; Larsen, B. R.; et al. Source attribution of urban smogepisodes caused by coal combustion. Atmos. Res. 2008, 88, 294–308.

(10) Lee, E.; Chan, C. K.; Paatero, P. Application of positive matrixfactorization in source apportionment of particulate pollutantsin Hong Kong. Atmos. Environ. 1999, 33, 3201–3212.

(11) Sawicki, E.; Meeker, J. E.; Morgan, M. J. The quantitativecomposition of air pollution source effluents in terms of aza-heterocyclic compounds and polynuclear aromatic hydrocar-bons. Int. J. Air. Water. Pollut. 1965, 9, 291–298.

(12) Dong, M. W.; Locke, D. C.; Hoffmann, D. Characterization ofazaarenes in basic organic portion of suspended PM. Environ.Sci. Technol. 1977, 11, 612–618.

(13) Nielsen, T. P.; Clausen, S.; Jensen, F. P. Determination of basicazaarenes and polynuclear aromatic hydrocarbons in airborneparticulate matter. Anal. Chim. Acta 1986, 187, 223–231.

(14) Chen, H. Y.; Preston, M. R. Azaarens in the aerosol of an urbanatmosphere. Environ. Sci. Technol. 1998, 32, 577–583.

(15) EU Directive 1999/30/EC of the European Parliament and ofthe Council relating to limit values for SO2, NOx, PM, and Pbin ambient air. Off. J. Eur. Commun. June 29, 1999, L 163/41/60.

(16) Chow, J. C.; Watson, J. G.; Lowentha, D. H.; Chen, L. W. A.;Zielinska, B.; Rinehart, L. R.; et al. Evaluation of organic markersfor chemical mass balance source apportionment at the FresnoSupersite. Atmos. Chem. Phys. Discuss. 2006, 6, 10341–10372.

(17) Fujita, E. M.; Zielinska, B.; Campbell, D. E.; Arnott, W. P.; Sagebiel,J. C.; Mazzoleni, L. ; et al. Variations in Speciated Emissionsfrom Spark-Ignition and Compression-Ignition Motor Vehiclesin California’s South Coast Air Basin. J. Air Waste Manage. Assoc.2007, 57, 705–720.

(18) Naumova, Y. Y.; Eisenreich, S. J.; Turpin, B. J,; Weisel, C. P.;Morandi, M. T.; Colome, S. D.; Totten, L. A.; Stock, T. H.; Winer,A. M.; Alimokhtari, S.; Kwon, J.; Shendell, D.; Jones, J.; Maberti,S.; Wall, S. J. Polycyclic Aromatic Hydrocarbons in the Indoorand Outdoor Air of Three Cities in the U.S. Environ. Sci. Technol.2002, 36, 2552–2559.

(19) Gehrig, R.; Hueglin, C.; Devos, W.; Hofer, P.; Kobler, J.; Stahel,W. A.; et al. Contribution of road traffic to ambient fine particleconcentrations (PM10) in Switzerland. Int. J. Vehicle Design 2001,27, 56–64.

(20) Putaud, J. P.; Raes, F.; van Dingenen, R.; Bruggemann, E.;Facchini, M. C.; Decesari, S.; et al. A European aerosolphenomenology - 2: Chemical characteristics of PM at kerbsite,urban, rural and background sites in Europe. Atmos. Environ.2004, 38, 2579–2595.

(21) Lanz, V. A.; Alfarra, M. R.; Baltensperger, U.; Buchmann, B.;Hueglin, C.; Prevot, A. S. H. Source apportionment of submicronorganic aerosols at an urban site by factor analytical modelingof aerosol mass spectra. Atmos. Chem. Phys. 2007, 7, 1503–1522.

(22) Kleeman, M. J.; Ying, Q.; Lu, J.; Mysliwiec, M. J.; Griffin, R. J.;Chen, J.; et al. Source apportionment of secondary organicaerosol during a severe photochemical smog episode. Atmos.Environ. 2007, 41, 576–591.

(23) Querol, X.; Viana, M.; Alastuey, F.; Amato, T.; Moreno, S.; Castillo,J.; et al. Source origin of trace elements in PM from regionalbackground, urban and industrial sites of Spain. Atmos. Environ.2007, 41, 7219–7231.

(24) Lanz, V. A.; Alfarra, M. R.; Baltensperger, U.; Buchmann, B.;Hueglin, C.; Szidat, S.; Wehrli, M. N.; Wacker, L.; Weimer, S.;Caseiro, A.; Puxbaum, H.; Prevot, A. S. H. Source Attribution ofSubmicron Organic Aerosols during Wintertime Inversions byAdvanced Factor Analysis of Aerosol Mass Spectra. Environ.Sci. Technol. 2008, 42, 214–220.

(25) Simcik, M. F.; Steven, J.; Eisenreich, S. J.; Lioy, P. J. Sourceapportionment and source/sink relationships of PAHs in thecoastal atmosphere of Chicago and Lake Michigan. Atmos.Environ. 1999, 33, 5071–5079.

ES8032082

7970 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 20, 2009