Atmospheric Research 143 (2014) 279–292

Contents lists available at ScienceDirect

Atmospheric Research

j ourna l homepage: www.e lsev ie r .com/ locate /atmos

Urban background levels of particle number concentration andsources in Vilnius, Lithuania

Steigvilė Byčenkienė, Kristina Plauškaitė, Vadimas Dudoitis, Vidmantas Ulevicius ⁎Center for Physical Sciences and Technology, Savanorių 231, 02300 Vilnius, Lithuania

a r t i c l e i n f o

⁎ Corresponding author. Tel.: +370 68635455, fax:E-mail address: ulevicv@ktl.mii.lt (V. Ulevicius).

http://dx.doi.org/10.1016/j.atmosres.2014.02.0190169-8095/© 2014 Elsevier B.V. All rights reserved.

a b s t r a c t

Article history:Received 26 July 2013Received in revised form 21 January 2014Accepted 14 February 2014Available online 4 March 2014

This study presents results of research on urban aerosol particles with a focus on the aerosolparticle number concentration (PNC) and the particle size distribution. The real timemeasurements of aerosol PNC (N4.5 nm) and number size distributions (9–840 nm) wereperformed. The seasonal variations essentially comprised theminimummonthlymean inOctober2010 (3400 ± 3000 cm−3) and themaximum inApril 2011 (19,000 ± 15,000 cm−3). Themeanannual PNC was 10,000 ± 8000 cm−3 with an average mode size of 30–50 nm. The presence ofstrong diurnal patterns in aerosol PNC was evident as a direct effect of three sources of aerosolparticles (nucleation, traffic, and residential heating). Hybrid receptor modeling potential sourcecontribution function (PSCF) and concentration weighted trajectory (CWT) were used byincorporating 72-h backward trajectories and measurements of PNC in Vilnius. The results oftrajectory clustering and the PSCF method demonstrated that possible additional source areascontributing to the elevated particle number concentration in Vilnius could be industrial areas incentral Europe. Principal component analysis (PCA) revealed highest loadings for PNC, PM10,NOx, NO, NO2 and SO2 concentrations, indicating combustion processes occurring in vehicleengines and use of sulfur-containing fossil fuels for residential heating.

© 2014 Elsevier B.V. All rights reserved.

Keywords:Particle number concentrationParticle size distributionPSCFCWT

1. Introduction

The adverse impacts of nanoparticles on the humanhealth (Brook et al., 2010) in urban environment havestimulated the scientific community to research urban airquality. The European Union, in common with many otheradministrations, has established tough targets for airquality in order to limit adverse effects upon human health.Although the mass of airborne particulate matter (PM) inthe ambient atmosphere is the subject to regulation, therecontinues a debate about which PM size fractions have amore harmful effect on human health. This is importantfrom a public health point of view because recent research

+370 52601723.

has indicated that ultrafine particles (UFP: particles withdiameters b100 nm) may be more toxic per unit mass thanthe larger size fractions of PM (Sioutas et al., 2005; Pope etal., 2002) and they sooner deposit in the deepest alveolarportions of the respiratory tract as a function of size (Daigleet al., 2003), induce oxidative stress (Araujo et al.,2008), and translocate into secondary organs and tissues(Oberdörster et al., 2004). The European Union (EU) set twolimit values for PM10 concentrations: annual average —

40 μg m−3, 24‐hour limit value — 50 μg m−3 not to beexceededmore than 35 times a calendar year (2008/EC/50).However, the implementation of the EU air-quality direc-tives and introduction of more stringent vehicle emissionstandards were deferred since some countries were unableto meet the pre-existing limit value. In consequence fact,the decreased PM mass concentration has increased theconcentration of UFP in Western countries (Morawska etal., 2002). Unavoidably, the legislative pressures arising

280 S. Byčenkienė et al. / Atmospheric Research 143 (2014) 279–292

from air quality standards turn attention to identification ofPM source categories and their geographic distribution.

Previous studies have suggested that the major influenceon contributions to the PNC is caused by the included vehicleexhaust emissions during the traffic peak hours and photo-chemical nucleation events (Ketzel et al., 2004; Holmes et al.,2005; Jeong et al., 2006; Cyrys et al., 2008; Pey et al., 2008;Perez et al., 2010), the residential heating during winterperiods (Hussein et al., 2004) and new particle formation(NPF) by photochemical reactions (Perez et al., 2010).Typically, the formation process occurs over large areas(Plauškaitė et al, 2010). Therefore, it may significantlycontribute to changing the regional climate (Spracklen etal., 2008).

In recent years, these events have been studied exten-sively in background (O'Dowd et al., 2002; Rodriguez et al.,2005) and also urban (or polluted) areas (Tuch et al., 2006;Kulmala et al., 2005). It was found that the anthropogenicemission sources such as manufactories and the powerplants (Köhler et al., 2008), the tire and road surface wear(Stocker and Carruthers, 2007), burning biofuel (Karvosenojaet al., 2008; Kumar et al., 2010), the aircraft activity (Hu et al.,2009), the natural sources (Holmes, 2007), the biomassburning (Simmonds et al., 2005; Wardoyo et al., 2007) andthe long-range transport (Beverland et al., 2000) wereimportant contributors to the PNC mostly in the b300 nmsize range (Kumar et al., 2008). The particles from theresidential combustion can sometimes grow in size in theurban environmentsmainly due to condensation of gases (Parket al., 2008). The exhaust emissions from the gasoline anddiesel-fuel vehicles remain the dominant source in pollutedurban environments (Harrison et al., 2011; Kumar et al., 2011).The size range for the road traffic emissions is consistent withthe particle number size distributions for the gasoline directinjection and the diesel engines with the majority of particlesin the size diameter interval of 20–60 and 20–130 nm,respectively (Maricq et al., 1998; Morawska et al., 1998,2008). These can alone contribute up to about 90% of thetotal PNCs (Perez et al., 2010) reaching magnitudes of 104–105 cm−3 during the nucleation events (Cheung et al., 2010).

In the case of the Lithuania, NPF has been studied byPlauškaitė et al. (2010) in a relatively clean backgroundmarine area site. These studies revealed NPF taking placemore frequently in conjunction with high levels of solarradiation. In spite of its importance, no long-term contin-uous measurement campaigns studying the PNC dynamicshave been carried out at urban sites in Lithuania so far. AsNPF was found to be a regional phenomenon it wasexpected that NPF would also occur in the polluted urbanareas. The current work is aimed at studying the sourceareas and processes affecting or contributing to PNC at anurban site for the period June 2010–September 2011. Toreach this objective, the work focuses on the seasonalevolution of the source apportionment of PNC based onback-trajectories clustering techniques to identify theregional area sources by the concentration weightedtrajectory (CWT) and the potential source contributionfunction (PSCF). Additionally some nucleation events havebeen identified and characterized to find out that the mostlikely conditions lead to high PNC concentration appear-ance in the measuring site.

2. Data set and methodology

2.1. Description of the study region

The investigations of particle size distribution and PNC arepresented within this paper. The aerosol PNC was analyzedduring the period from 1 June 2010 to 30 September 2011 inVilnius. The continuous aerosol measurements were taken atthe top floor of the academic building of the Center forPhysical Sciences and Technology campus located in Vilnius.The inlet of sampling system was placed on the top floorabout 20 m above the ground level, 12 km southwest ofdowntown area. The location can be described as an urbanbackground (Fig. 1).

Under the normal meteorological conditions, the poten-tial for accumulation of the vehicle emissions is limited bythe site location. The urban sampling site was relatively farfrom the local sources of the primary particles. The nearesthighway was 0.24 km to the southwest; on the opposite sidea low traffic road was 0.6 km away.

2.2. Instrumentation

The aerosol PNC was continuously measured using aCondensation Particle Counter (CPC, custom built) UF-02(Mordas et al., 2005). The CPC was designed to detect theparticle number concentration from 0.002 up to 100,000 cm−3.The detection efficiency reaches a value of 1 at large particle sizesand it is smaller than 0.9 for particles smaller than approximately5 nm. The maximum observable number concentration (accu-racy 20%) is 100,000 cm−3. The detection efficiencies of CPCwere carefully measured in the calibration set-up presented byMordas et al. (2005). The design of the instrument is basedon the swirling flow generated inside the saturator (43 °C)-condenser (10 °C). The instrument uses a high flow rate(1 l min−1) of the carriers. From the carrier flow, the aerosolflow (0.27 l min−1) is extracted by a capillary. This aerosolflow is divided into two flows. The first one (0.03 l min−1) isdirected to the condenser. The second flow (0.24 l min−1) iscirculated through a HEPA filter and a saturator block, in whichthe flow is saturated with respect to n-butanol and thenmixedwith the aerosol-laden air before cooled condenser. Thismixing generates a supersaturated region with respect ton-butanol. The n-butanol vapor condenses on the particleswhich act as condensation nuclei. This process increases thesize of each individual nanoparticle. Such large droplets can beconveniently detected by light scattering. The lower cut-off sizeof the CPC, i.e. the limiting size when 50% of the particles aresuccessfully accounted for, is determined to be ≥4.5 nm(Mordas et al., 2005). The instrument is fittedwith an impactor(laminar flow, nozzle diameter = 7.4 mm) to reduce theinfluence of the large particles. The yearly maintenanceincluded the CPC calibration and thorough cleaning.

A Magee Scientific Company Aethalometer™, Model AE40Spectrum,manufactured by Optotek, Slovenia was deployed atthe Vilnius site and provided realtime, continuous measure-ments of the BC mass concentrations. The optical transmissionof carbonaceous aerosol particleswasmeasured sequentially atseven wavelengths λ (0.37, 0.45, 0.52, 0.59, 0.66, 0.88 and0.95 μm). The 0.88 μm is considered as the standard channelfor BC measurements as at this wavelength BC is the principal

Fig. 1. Location of the monitors in the city of Vilnius, Lithuania.

281S. Byčenkienė et al. / Atmospheric Research 143 (2014) 279–292

absorber of light (Lavanchy et al., 1999). The aethalometeroutput is calculated directly as the BC concentrations throughan internal conversion using assumed mass absorption effi-ciency. The aethalometer converts light attenuation to the BCmass using a fixed specific attenuation cross-section (σ) of16.6 m2 g−1 of BC (Aethalometer™Operationsmanual, MageeScientific). This is the default value set by the manufacturer fora wavelength of 880 nm. The aethalometer data recorded witha 5-minute time base were compensated for loading effectsusing an empirical algorithm (Virkkula et al., 2007). Theaethalometer was equipped with an additional impactorremoving the particles larger than 2.5 μm of the particleaerodynamic diameter. The starting time referred in thispaper is Greenwich Mean Time (for local time: GMT + 2:00).The measurement precision of the Aethalometer is reported tobe ±100 ng BC m−3 with 1-minute average at a flow rate of150 ml min−1 as specified in technical specifications by themanufacturer. It is sufficient for ambient total BC concentrationmeasurement with typical range (1–10 μg m−3) in urbanenvironments.

Other data including PM10 and mass concentrations ofPM2.5, NO, NOx, NO2, meteorological factors (temperature,relative humidity, wind speed, and wind direction (T, RH, WS,and WD, respectively)), and gaseous pollutants (SO2, CO, O3)were supplied by two stations (Savanorių and Lazdynai)belonging to the Environmental Protection Agency (EPA) lessthan 1 km from our monitoring site (http://stoteles.gamta.lt/).PM10 and PM2.5 measurements were conducted with theEnvironnement S.A.MP101Mmonitor (based onβ-attenuation).Continuous trace gas measurements of NOx (NO and NO2), SO2,CO, and ozone were made using chemiluminescence for NOx

(Environnement S.A., Model AC31M), UV fluorescence for SO2

(Environnement S.A., Model AF21M), infrared gas filtercorrelation for CO (Environnement S.A., Model CO11), and UVabsorption (Environnement S.A., Model O341M) for ozone,respectively.

Aerosol particle number size distribution was obtained bythe use of a Scanning Mobility Particle Sizer (SMPS; from 9 to840 nm) based on the principal of the mobility of a chargedparticle in the electric field. The SMPS system consists of asingle type differential mobility analyzer (DMA) (LeibnizInstitute for Tropospheric Research) and CPC UF-02 anaerosol neutralizer, a control unit and data logging system.A scan time of 180 s and retrace time of 120 s were employedfor each sample. The sheath flowwithin the system is 5 l/minand the sample flow rate — 1 l/min. The relative humidity ofthe sample does not exceed 50%. The aerosol particle numbersize distribution values were measured at 1 min intervalsand the particle size distribution determined at 5 minintervals. These measurements are averaged on and groupedon an hourly basis, and thereby reported as an hourlyaverage. Prior to measurements, the relative humidity (RH)of aerosol sample was reduced to 30% inside a Nafion drier.

2.3. Statistical investigation

2.3.1. Principal component analysisPCA is a special case of factor analysis that transforms the

original set of intercorrelated variables into a set ofuncorrelated variables. PCA is a method that helps extractmore information from a time-series than when individualparameter analysis is used (Fahrmeir et al., 1996; Einax et al.,1997). It extracts the directions in which a cloud of datapoints is maximally stretched, i.e. has a maximal variance.

282 S. Byčenkienė et al. / Atmospheric Research 143 (2014) 279–292

The most relevant information of the data set (J variableswith K observations) is contained in these directions (i.e.principal components (PCs)). The PCs represent orthogonaland therefore independent linear combinations PCi of the Joriginal variables v

PC ið Þ ¼XJ−1

j¼1

bij � vj; ð5Þ

where bij are the component loadings and indicate howstrongly a specific original variable vj contributes to PCi, and vjis the original variable. PCs are found by calculating theeigenvectors and eigenvalues of the data covariance matrix.The projection of the original data on the eigenvectorsdefines the PCs, and the eigenvalue of every eigenvectorindicates the contribution of the specific PC to the total dataset variance. There are several equivalent ways of derivingthe principal components mathematically. The simplest oneis by finding the projections which maximize the variance.The first PC is the direction in feature space along whichprojections have the largest variance. The PC 2 is thedirection which maximizes variance among all directionsorthogonal to the first. The PC 1 carries most of theinformation about the data (i.e. explains most of the variancein the data), the PC 2 one will then carry the maximumresidual information, and so on.

2.3.2. Trajectory cluster analysisThree-dimensional 72-h back trajectories of air masses

arriving at Vilnius site (i.e., the receptor site) were calculatedusing the Web version of the Hybrid Single ParticleLagrangian Integrated Trajectory (HYSPLIT-4) model anddata from NCEP/NCAR Reanalysis meteorological database(Draxler, R.R. and Rolph, G.D., 2014). The computationstarted from the Vilnius site for identifying the air masspaths and the origin of air masses, which could have an effecton PNC. Transport paths of air masses with similar historyand origin were obtained by cluster analysis. Using a clusteralgorithm, the homogeneity within clusters was achieved byminimizing the angle distances (Sirois and Bottenheim,1995) between the corresponding coordinates of the indi-vidual trajectories (considering the full length of each 72-hair mass backward trajectory).

The angle distance between two backward trajectorieswas then given by:

d12 ¼ 1n

Xni¼1

cos−1 0:5A1 þ Bi−Cið Þffiffiffiffiffiffiffiffiffi

AiBi

p !

; ð2Þ

where

Ai ¼ X1 ið Þ−X0ð Þ2 þ Y1 ið Þ−Y0ð Þ2Bi ¼ X2 ið Þ−X0ð Þ2 þ Y2 ið Þ−Y0ð Þ2Ci ¼ X2 ið Þ−X1ð Þ2 þ Y2 ið Þ−Y1ð Þ2:

ð3Þ

The variables X0 and Y0 define the position of the studiedsite X1(i), Y1(i) and X2(i), Y2(i) are coordinates of i segmentfor trajectories 1 and 2. In order to investigate the possibleseasonal variation in transportation process of PNC, weanalyzed all four seasons.

The final cluster centers were computed as the mean foreach variable within each final cluster. Since each backwardtrajectory is associated with a set of PNC measured at Vilnius,clusters are associated with average PNC.

2.3.3. PSCF and CWT methodsPSCF and CWT were used to locate regional source areas

potentially affecting PNC level at Vilnius. The PSCF technique,developed by Ashbaugh et al. (1985) is a conditionalprobability function giving the probability that an air parcelwith a certain level of pollutant concentration arrives at areceptor site after passing through a specific upwind sourcearea (Hopke and Hwang, 2007). To calculate the PSCF, thewhole European region covered by the trajectories wasdivided into a gridded i by j array. In this study the gridcovers area of interest defined by (40–70) N and 20 W–40Ewith the center of Vilnius site (54.41°N, 25.17°E) as themidpoint, and contained grid cells of 0.5° × 0.5°. The arrivalheight was chosen as 100 m above ground level to representa well-mixed convective boundary layer for regional trans-port investigation. This height was chosen to weaken theeffects of surface friction. A starting height of 100 m has beenused in a number of prior publications (Liu et al., 2013;Riuttanen et al., 2013).

Mathematically, the PSCF is a function of location asdefined by the cell indices i and j while the number ofsegments with endpoints that fall in the ijth cell is denoted bynij. The number of endpoints in the ijth cell associated with atrajectory that arrives at the sampling site at the same time asa corresponding measured pollutant concentration higherthan an arbitrary criterion value is defined by mij. The PSCFvalue for the ijth cell is then:

PSCFij ¼ mij=nij ð4Þ

where nij is the number of trajectories that originate in theijth grid during the study period and mij is the number oftrajectories that arrived at a receptor site with pollutantconcentrations higher than a specified criterion value. It isimportant to note that a grid with no end points (nij = 0)cannot be identified as a source area in the analysis eventhough there are known emission sources in the grid cell(Cheng et al., 1993).

Then the value of PSCF was interpreted as the probabilitythat the concentration of a PNC greater than the creationlevel is related to the passage of air parcel through the ijthcell. These cells are indicative of areas of high potentialcontributions for that pollutant.

CWT is a function of PNC concentrations that werereported every 3-h and the residence time of a trajectoryarriving at Vilnius in each grid cell. For every concentrationone air mass backward trajectory was generated. Thus, it isdifferent from PSCF, which focuses on the residence time of abackward trajectory in each grid cell only above an arbitraryconcentration threshold. The CWT model parameters select-ed were the NCEP/NCAR Reanalysis archive meteorologicaldata from the NWS NCEP, trajectory duration of 72-h, andstarting height of 100 m. The hourly trajectory segmentendpoints for each back trajectory that corresponds to each1 h PNC were retained. For 72-h trajectory duration, therewere normally 72 trajectory segment endpoints.

283S. Byčenkienė et al. / Atmospheric Research 143 (2014) 279–292

This domain encompasses the locations of nearly all thetrajectory segment endpoints from all the back trajectoriesand PNC sources. The geographical domain was divided into16,000 grid cells (200 grid cells across the latitude by 80 gridcells across the longitude), each covering an area of0.5° × 0.5°. The CWT is a measure of the source strength ofa grid cell to the Vilnius site and is determined as follows(Jeong et al., 2011; Kabashnikov et al., 2011):

CWTi; j ¼

XLT¼1

CTτi; j;T

XLT¼1

τi; j;T

: ð5Þ

CT is the 1 h PNC concentration corresponding to thearrival of back trajectory T; τi,j,T is the number of trajectorysegment endpoints in grid cell (i,j) for back trajectory Tdivided by the total number of trajectory segment endpointsfor back trajectory T; and L is the total number of backtrajectories over a time period (i.e. each season). Given CT forPNC, τi,j,T can be determined by counting the number ofhourly trajectory segment end-points from the HYSPLITmodel in each grid cell for each trajectory. This was repeatedfor all the back trajectories, L.

PSCF and CWT were determined for each season (Winter:December, January, February; Spring: March, April, May;Summer: June, July, August; Autumn: September, October,November).

3. Results and discussion

In this study particle number concentration and dynamicsin the urban environment over thirteenmonths (1 June 2010 to30 September 2011) are provided (See Section 3.1). Thedetection of relationships among the measured parametersrequires an in-depth analysis that includes statistical methodssuch as correlation analysis and principal component analysis(see Section 3.2). Air mass backward trajectory clustering,potential source contribution function and concentrationweighted trajectory analyses were used to determine the

Table 1Statistical analysis of aerosol PNC, (cm−3).

Year Month N Mean SD

2010 Jun 530a 8400 6000Jul 327a 7600 5000Aug 169a 8000 5000Sep 385a 12,000 8000Oct 87a 3400 3000Nov – – –

Dec – – –

2011 Jan 745 16,000 14,000Feb 673 16,000 14,000Mar 745 15,000 13,000Apr 720 19,000 15,000May 745 14,000 13,000Jun 698 8000 6000Jul 665 4000 3000Aug 745 3500 3000Sep 721 7100 7000

a Missing data (N15%) due to instrument failure.

most probable geographical location and origin of aerosolparticles (Section 3.3). Three of the most representative typesof days (7, 18 and 22 September 2011) in terms of diurnalparticle size distribution and number concentration behaviorwere investigated.

3.1. Overview of the particle number concentration

3.1.1. Seasonal and diurnal variationsThe aerosol particle number concentration levels mea-

sured in Vilnius displayed a large variability mainly related todaily activities and season. However, a mean concentrationlevel of the order of 10,000 ± 8000 cm−3 was recognized asthe urban background. The descriptive statistics for the 1-hmean aerosol PNC are presented in Table 1.

There is one major data gap because CPC was broken inNovember 2010 and instrument was again re-located inJanuary 2011. During the entire study period there wereseveral pronounced increases in the aerosol PNC. It followsfrom Table 1 that the seasonal variations essentially comprisethe winter–spring peak, which is noticeable during themonth of April. Similar concentrations covering the wholecontinent were detected annually (Byčenkienė et al., 2011);the monthly mean aerosol PNC varied in 2010 from 3400 ±3000 (October) to 12,000 ± 8000 cm−3 (September) andfrom 3500 ± 3000 (August) to 19,000 ± 15,000 cm−3

(April) in 2011. The maximum hourly concentration duringthe whole measurement period was observed in February2011 (113,000 cm−3) and the minimum in June 2010(100 cm−3). It is suggested that without the sufficientphotochemical processes under the winter conditions, theweak convection in the absence of high temperature andstronger emission of primary particles during residentialheating period may cause the accumulation of nano-particles(Gao et al., 2007). The higher aerosol PNC was usuallyobserved during the cold season, when the air temperaturedropped below zero. This is explained by the domestic use offossil fuels and biofuels for residential heating. The recentstudies by Karvosenoja et al. (2008) and Meyer (2012)indicate that the domestic use of the fossil and wood fuelscould be a major source of aerosol particles. Also, the peak of

P25 P75 P95 Min Max

4000 11,000 20,000 100 49,0004000 10,000 15,000 2000 40,0005000 10,000 16,000 2000 33,0006000 15,000 27,000 300 61,0001000 5000 9000 400 18,000– – – – –

– – – – –

6000 20,000 42,000 1000 92,0007000 19,000 39,000 1000 110,0008000 18,000 39,000 1000 100,0009000 24,000 51,000 3000 110,0006000 17,000 41,000 2000 73,0003000 11,000 21,000 100 46,0002000 5000 10,000 800 24,0002000 4000 9000 400 22,0003000 8000 21,000 200 59,000

Fig. 2. Weekday variations of PNC (Vilnius) over the study area. Box coversinterquartile range between 25th and 75th percentiles. The square in thebox interior represents the mean, the horizontal line in the box interiorrepresents the median, the vertical lines issuing from the box denote the 5thand 95th percentile values.

284 S. Byčenkienė et al. / Atmospheric Research 143 (2014) 279–292

PNC related to the long-range or regionally transportedsmoke emitted by the biomass burning (Ulevicius et al.,2010; Byčenkienė et al., 2011) had a strong impact on thetotal aerosol number concentrations in Lithuania over springperiod. In comparison, Jaenicke (1993) gives the total PNC ofabout 6000 cm−3 for the remote continental areas, whereasour values in Preila were about 2600 ± 900 cm−3 in2008–2009 and 10,000 ± 8000 cm−3 in 2010–2011 inVilnius (urban background environment). Depending onseveral factors (e.g. traffic intensity, season, weather), thesevalues are in agreement with those reported in Brisbane,ranging between 7500 and 15,400 cm−3 (Morawska et al.,1998), but are lower than the average concentration ofaround 12,500 and 29,500 cm−3 reported for an urbanresidential area in Essen, Germany, during summer 2008(Weber, 2009). Concentration levels in this study are similarto those measured at an urban street station in centralBudapest (6–1000 nm, 11,800 cm−3) from 3 November 2008to 2 November 2009 (Salma et al., 2011) and in Finland atSiltavuori urban background site (8–400 nm, 18,040 cm−3)(Laakso et al., 2003) in winter as well as experiment inGwangju (7400 cm−3) (Park et al., 2007), in substantialagreement with data reported for Mediterranean urban

Table 2Spearman correlation coefficients between pollutant concentrations and meteorolo

PM10 SO2 NO NO2 NOx O3

PM10 1 0.17 0.32⁎ 0.62⁎⁎ 0.61⁎⁎ −0.62⁎⁎

SO2 1 0.1 0.29⁎ 0.26 −0.12NO 1 0.35⁎ 0.71⁎⁎ −0.35⁎

NO2 1 0.91⁎⁎ −0.89⁎⁎

NOx 1 −0.82⁎⁎

O3 1RHpWSCOPNC

⁎ Correlation is significant at the 0.05 level.⁎⁎ Correlation is significant at the 0.01 level.

centers (10,200, 11,000 and 10,300 cm−3) such as Augsburg,Helsinki and Stockholm, respectively (Aalto et al., 2005).

The weekly pattern of aerosol PNC averaged over theentire period is shown in Fig. 2. Measurements have shownthat the highest aerosol PNC mean detected in Vilnius was13,000 ± 8000 and 12,500 ± 8000 cm−3 (on Tuesday andFriday, respectively).

Slightly lower concentration on Monday than on Tuesdaycan represent lower emissions on Sunday in the pollutionsource regions. The clear influence of the road traffic intensityon the PNC increased levels was observed for workdays12,000 ± 8000 cm−3 and 8400 ± 8000 cm−3 on weekends,respectively. The lowest concentrations in the 75th percen-tile occurred on Saturday (8000 cm−3).

3.2. Statistical investigation

3.2.1. Pollutant concentration correlation analysisVenter et al. (2012) indicated that NOx, CO, BC and PM10

mainly originated from the local residential heating andtraffic, while SO2 generally originated from the use ofsulfur-containing fossil fuels for domestic heating. TheSpearman correlation coefficients among PNC, meteorologi-cal parameters and trace gases are presented in Table 2.Higher wind speed leads to a decrease of all pollutants(except O3). Thus, the increased concentrations are enhancedby low wind speeds. The Spearman correlation coefficientsbetween NOx, NO and CO were moderately good withcoefficients mostly in the range of 0.5. Because CO is emittedfrom gasoline powered vehicles, moderate correlation wasexpected with these pollutants, while poor correlationbetween CO and other pollutants (PM10, SO2, PNC) wasobserved as they dominated by diesel vehicles. Similar lowcorrelations were reported by Bukowiecki et al. (2002)between CO and ultrafine particle number concentrations.Good correlation is found with PNC and PM10 (r = 0.58).Nevertheless, statistically significant but weak correlationsare found between PNC and NO2 (r = 0.31).

As shown in Table 2, NOx (r = 0.46) and O3 (r = 0.41)were moderately correlated with the particle numberconcentrations while SO2 and CO levels were found to bepoorly correlated with the daily PNC. From the resultsreported in Table 2 it can be argued that the high correlationsamong PNC, PM10 and NO are in agreement with theinfluence of anthropogenic emissions due to the road traffic

gical parameters.

RH p WS CO PNC

0.131 −0.20 −0.56⁎⁎ 0.26 0.58⁎⁎

−0.33⁎ 0.06 −0.26 −0.27 0.02−0.19 −0.39⁎⁎ −0.24 0.28 0.51⁎⁎

0.14 −0.24 −0.81⁎⁎ 0.49⁎⁎ 0.31⁎

0.02 −0.35⁎ −0.72⁎⁎ 0.45⁎⁎ 0.46⁎⁎

−0.37⁎⁎ 0.45⁎⁎ 0.64⁎⁎ −0.56⁎⁎ −0.41⁎⁎

1 −0.27 −0.10 0.36⁎ −0.051 0.21 −0.51⁎⁎ −0.35⁎

1 −0.43⁎⁎ −0.181 0.15

1

Table 3Principal component matrix.

Components

1 2 3

PM10 0.738 0.154 −0.023SO2 0.185 0.734 −0.286NO 0.586 0.234 0.599NO2 0.902 0.109 −0.335NOx 0.941 0.186 0.010O3 −0.902 0.157 0.176RH 0.201 −0.770 −0.270p −0.505 0.387 −0.414WS −0.766 −0.116 0.415CO 0.616 −0.555 0.020PNC 0.550 0.155 0.586Cumulative, % 45 62 74

285S. Byčenkienė et al. / Atmospheric Research 143 (2014) 279–292

(Marconi et al., 2007). Moreover, in a recent study, poorcorrelation (r = 0.15) at urban background level wasobserved between CO and PNC, indicating that traffic is notthe major source of PNC in the urban background area(Wahlin et al., 2001). On the other hand, the high correlationfound between such gaseous compound as NO and PNCagrees essentially with other studies carried out in otherEuropean cities, in which it was shown that NO can beconsidered a better tracer of traffic related pollutants, thantraffic intensity itself (Ruuskanen et al., 2001; Janhäll et al.,2004). The significant correlation between PNC and NO

Fig. 3. Trajectories representing grouping of 72-h backward trajectories of air masseasons. Seasonal variations of the potential source maps for PNC arriving at 100 m

concentrations may suggest that nucleation occurs as thevehicle exhaust mixes with the cool ambient air (Shi et al.,1999).

3.2.2. Principal component analysisIn order to identify the sources contributing ultrafine

particles, principal component analysis was applied to thedata using Kaiser normalization and Varimax rotated methodin SPSS statistical software packages (SPSS Inc., USA).

Table 3 lists the key results of the calculated PCA in termsof the (factor) loadings of the variables for the first three PCs.The last row in the table gives the percentage of the variancein the data that can be explained by the respective PC and thehigher level PCs. To interpret the results, the first threeindividual PCs were evaluated. PC 1 explains 45% of thevariance in the data, and PC 2 and PC 3 explain 62 and 74% ofthe variance, respectively. PCs 4 to 11 are not shown becauseof their relatively low explained variance.

The PCA revealed the highest loadings for PNC, PM10,NOx, NO, NO2 and SO2 concentrations, indicating combustionprocesses occurring in vehicle engines and the use ofsulfur-containing fossil fuels for residential heating. PC 1represents emissions from other combustion processes suchas residencial heating. In addition, PCA also revealed that PNCdepended on meteorological conditions such as wind speedand pressure. The negative correlation (−0.902) with the O3

concentration can be explained by the chemical balancebetween NOx and O3.

ses over Vilnius into six classes for the winter, spring, summer and autumnaltitude in Vilnius: (A) winter, (B) spring, (C) summer, and (D) autumn.

Fig. 4. CWT map for PNC (cm−3) during different seasons.

286 S. Byčenkienė et al. / Atmospheric Research 143 (2014) 279–292

The second PC explained 17% of the variance and has highloadings for SO2 and negative loadings of CO and RH. It isknown that SO2 represents residential heating while COexists in vehicular emissions. The third PC was found to havehigh positive loading for NO which could be emitted by cartraffic as nitrogen is recognized as a suitable tracer of thetraffic emissions in the urban environments and, moreover,was reported as highly influenced by the photochemicallyinduced nucleation (Fernández-Camacho et al, 2010; Cheunget al., 2010).

3.3. Source regions

3.3.1. Cluster analyses of air mass back trajectories, potentialsource contribution function and concentration weightedtrajectory analysis

PSCF model, CWT method and cluster analysis were runwith the seasonal data for winter, spring, summer andautumn in order to identify the main atmospheric circulationpathways influencing PNC (Figs. 3–4). There are fourdominant paths of air masses reaching Lithuania: from theW, NW, SW and SE, as shown in Fig. 3. The long and fastmoving trajectories were disaggregated into groups originat-ing from more distant W and NW regions. Members of thiscluster have extremely long transport patterns; some of themcross over North Sea, Northern Europe or the Nordiccountries. Trajectories belonging to S–SW typically follow aflow pattern over the Poland and Germany. Generally suchtrajectories have short transport patterns, indicative ofslow-moving air masses. Most of the strong PNC episodes

within this group are probably enriched from regional andlocal emission sources. Southern group of trajectories followsone of the altitudes closest to the earth surface (typicallyunder 1000 m). Eastern Europe cluster trajectories weresignificantly lower than those from Western Europe. More-over, Eastern clusters are characterized by the shortesttrajectories, which are mainly transported from locationsbetween E and SE. A summary of essential cluster character-istics is given in Table 4. The polluted air mass trajectories arethose with PNC higher than the mean in appropriate season.

In spring no clusters exceed the spring season mean of16,300 cm−3, but cluster 1 (26.1%) was recognized aspolluted one, which indicates that the air masses associatedwith this cluster lead to higher PNC loadings (12,700 ±10,300 cm−3) in Vilnius. About 20.7% of trajectories wereassigned to cluster 3 with the lowest concentration of5500 ± 12,200 cm−3, thus this cluster had less effect onPNC loadings in Vilnius. Cluster 2 (16.7% of trajectories)represents the most important transport pathway with thehighest PNC (15,500 ± 8600 cm−3) among the six clustersin winter. The second important transport pathway wasassociated with cluster 6 (6.8%), which had the second highPNC (12,300 ± 10,900 cm−3). The number of trajectoriesassigned to cluster 3 was high (35.6%), therefore, thepathway represented by this cluster was considered impor-tant for lowest PNC loadings (9900 ± 7600 cm−3). Inautumn, the pathway represented by cluster 4 was mostpolluted (11,200 ± 6300 cm−3), while those of clusters 5(4100 ± 7700 cm−3) and 6 (4000 ± 2500 cm−3) wererelatively clean.

Table 4Mean PNCs for all trajectory clusters arriving at Vilnius.

Season/cluster Spring Winter Summer Autumn

Mean PNC,cm−3

Stdev,cm−3

Mean PNC,cm−3

Stdev,cm−3

Mean PNC,cm−3

Stdev,cm−3

Mean PNC,cm−3

Stdev,cm−3

1 12,700 10,300 11,050 23,500 4100 3100 6900 51002 9000 9100 15,500 8600 3900 5700 5500 57003 5500 12,200 9900 7600 3800 2200 4400 11704 7400 6300 11,000 9100 2805 1200 11,200 63005 8300 10,400 12,100 9000 5200 7000 4100 77006 10,100 7700 12,300 10,900 4500 4600 4000 2500

287S. Byčenkienė et al. / Atmospheric Research 143 (2014) 279–292

The potential source maps for PNC arriving at 100 maltitude in Vilnius during study period for each season aregiven in Fig. 3. Fig. 3 clearly shows that airflows follow aseasonal pattern, since seasonally different PSCF results wereobtained. Central European countries were the greatest PNCcontributors to Lithuania atmosphere in 2010–2011. Cellswith high PSCF values (appeared over Poland, Germany andCzech Republic) were the potential source regions to haveeffect on high PNC in Vilnius. Some Russian regionsimmediately west of Belarus and Ukraine also had highPSCF values. Similarly to the 2004–2008 results reported byMijić et al. (2012) the highest average contribution to theobserved PM10 concentrations during the winter season wasfrom the air mass cluster representing the arrival directionfrom west Ukraine. Riuttanen et al. (2013) analyzed thesource areas for SO2 and NOx transported to Hyytiala during1996–2008. The most likely sources affecting concentrationlevel were found to be located mainly in Eastern Europe.

So we assumed that the reason for the high PSCF valuesmust be attributed to airflow loaded with PNC originatingfrom the previously mentioned countries (in winter) and NPF(in spring) (Fig. 3A, B). Recent findings indicate that cleanarctic and marine air masses have been shown to be optimalfor NPF in Finland (Hyytïala) and Lithuania (Preila) (DalMaso et al., 2007; Plauškaitė et al., 2010). The frequency of allnucleation events and the relative number concentration ofnewly formed nucleation mode particles were higher incleaner cold air masses from the North Atlantic in March–

Fig. 5. A–C. Daily variability of dN/dlo

April and September (Dal Maso et al., 2007). The summerPSCF plots clearly had more 0.9–1.0 factors (red-shaded)than those in winter and spring. But similar to other seasons,the trajectories in the summer traveled as far as northernPoland and Belarus. The whole territory of Lithuania and thepart of Baltic Sea also had high PSCF values. This result agreeswith that of Mijić et al. (2012), in which different transportpattern of PM10 during summer and winter season wasobtained.

Fig. 4 shows the distribution of weighted trajectoryconcentration which gives the information on the relativecontribution of source regions potentially affecting PNC atVilnius. CWT is a function of PNC that was reported every24 h and the residence time of a trajectory arriving at Vilniusin each grid cell. From this perspective, it is different fromPSCF, which focuses on the residence time of a backtrajectory in each grid cell only above an arbitrary concen-tration threshold.

The CWT values revealed that PNC observed in Vilnius isnot heavily affected by long-range transport of air massesduring summer as CWT values are less variable. There aresome trails approaching from the N (autumn) and NW(winter), indicating that regional transport from NorthernEurope was not always negligible. Eastern parts of Belarusand Ukraine were the regions with high CWT values(N12,000 cm−3) in autumn. The pollution accumulated inthe southern countries' air masses presents a contribution, orrather a baseline, to which local emissions (which are

gDpN and modal distributions.

288 S. Byčenkienė et al. / Atmospheric Research 143 (2014) 279–292

possibly dominant) are added. Most of reported winterepisodes in Europe were caused by long range transportfrom sources of particulate matter, such as coal/woodcombustion for heating (Juda-Rezler et al., 2011; Kukkonenet al., 2005), as well as by increased traffic emissions due tounfavorable winter driving conditions (Morawska et al.,1998, 2008). Wood burning along with domestic waste andpoorest and least expensive types of fuel is probably widelypresent in individual heating houses not only in Poland(Zwozdziak et al., 2012).

3.4. Representative types of NPF events

In this section we report three of the most representativetypes of days, in terms of diurnal particle number sizedistribution and aerosol particle number concentrationbehavior: (1) the days with the NPF (Perez et al., 2010),(2) in situ nucleation due to residential heating appliances(Hussein et al., 2004) and (3) rush hours (Holmes et al.,2005), since they influence particle concentration anddynamics of particle size distributions. The analysis ofcharacteristic such as geometric mean diameter duringcontinuous measurements allows the identification of theNPF or the high-polluted events observed comprises theidentification of NPF bursts, obtaining characteristic such asgeometric mean diameter (Dg).

When nanometric particles had to appear for at least anhour in the size distribution in the nucleation mode andprevails, it was assumed that the NPF took place. It wasrevealed that based on the distinct shape of the aerosolparticle formation and growth observed on the contour plotsof the number size distribution, as previously described byDal Maso et al. (2005), most NPF events during the studyperiod occurred during daytime. Fig. 5(A, C, E) shows thediurnal variations of particle size distribution with thediameter on the left y-axis, the normalized particle concen-tration (dN/dlogDp) on the right, and time of the day on thex-axis, and additional A box is added to this figure to indicatewith accentuating quadrangle for several distinct types ofultrafine particle events: (a) “10–20 nm traffic event” (07/09/2011), (b) “20–90 nm photochemical NPF event” (18/09/2011), (c) “10–30 nm traffic and 100–300 nm residentialheating event” (22/09/2011). Fig. 5(A, B, C) and Table 5 showthe modal distribution of the normalized particle numberconcentrations as well.

Typically, the maximum of the diurnal PNC variation maybe attributed to traffic emissions (Kittelson et al., 2000)where under favorable conditions nucleation can occur, forexample, in road tunnels or in open road areas (Kittelson andWatts, 2000). During the study period, the particle numbersize distributions were characterized by the maximum PNCin the Aitken or accumulation modes before the events(Fig. 5). Fig. 5 presents the NPF in the afternoon and evening(07/09/2011, 18/09/2011) followed by a clear growth toaccumulation mode indicated by the “banana shape” temporaldevelopment of the number size distribution. The nucleationstarted at about 13:00 (Fig. 5A) and 18:00 (Fig. 5B) and thePNC in the 10–20 nm size range increased sharply to about80,000 cm−3. The NPF event was proved to be an importantsource of the ~100 nm particles in this study, except duringrush hours. The anthropogenic, non-traffic related pollution

episodewas observed during 22 September 2011 (Fig. 5C). Thebimodal shape is assumed to indicate the presence of thesecondary aerosol particles formed during the photo-chemically active period, i.e. the sunrise maximum can beprolonged and the concentration increases due to theformation of new secondary particles. Fig. 5(A, B) demon-strates that both road traffic emissions and the NPF under thesame conditions can substantially contribute to the atmo-spheric PNC.

The geometric mean diameter and the PNC were used tocharacterize the source strength. During the residentialheating and traffic event day (Fig. 5C, Table 5) mode 1particle concentration was found to be 3.2 × 105 cm−3 withthe smallest Dg = 37.6 nm suggesting traffic influence; forpotential source of smoke trace from the local wood burningfor house residential heating mode 2 particle concentrationwas found to be NTot = 4.3 × 104 cm−3 and Dg = 137 nm,while for the NPF event day, the Dg of ultrafine particles wasfound to be 80 nm. From Fig. 5(C), one can note that thephotochemical NPF event provided the lowest average PNC inthe 20–90 nm size range (NTot = 1.5 × 105 cm−3) with anaverage of Dg = 33.2 nm. This is further confirmed bycoincidence with the average Dg = 25 nm for photochemical(2/8/2006), 48 nm for traffic (24/2/2007) events along withpoor agreement with 53 nm for residential heating (24/2/2007) events, respectively, reported by Park et al. (2008). TheDg values were found to be higher than those observed inurban Gwangju, Korea (for residential heating appliances andNPF) due to different size ranges of events and probablyhigher source importance such as traffic and residentialheating at this site. Watson et al. (2006) also showed that theDg for the traffic events was typically ~30 nm.

3.4.1. The diurnal patterns of aerosol PNC during eventsThe previously reported measurements at the traffic-

dominated sites determined that the maximum PNC occurredfor diameters around 20 nm (Wehner et al., 2009). Vehiclesdirectly emit a significant number of particles in the sizerange between 40 and 100 nm (Harris and Maricq, 2001).The results of analysis investigated with respect to thediurnal pattern of aerosol particle number concentration ofeach mode were considered. As it is seen in Fig. 6, the particlesize distribution on 7 September 2011 showed the nucleationmode particles with the diameter of about 9–20 nm.

It can be seen that a peak number concentration wasobserved on 7 September during the early morning andafternoon, while the PNC was relatively stable on 22September. The concentration of the nucleation mode wasmuch higher than Aitken and accumulation modes, indicat-ing the significant contribution of traffic pollution to thenumber concentration for both events. In Fig. 6, the numberconcentrations of nucleation mode (N10–20), Aitken mode(N20–90), and accumulation mode (N100–300) particles werecalculated. On “traffic event day” (Fig. 5A) the nucleationstarted at about 12:00 and N10–20 increased sharply to about50,000 cm−3. Then the newly formed particles grew tolarger sizes, so N100–300 increased a few hours later to10,000–20,000 cm−3. As new particles can grow to largersizes and affect the optical properties of the atmosphericaerosol particles (Allan et al., 2006) the dramatic change ofthe wavelength dependence of black carbon absorption

Table 5Aerosol particle size distribution parameters.

Curves A B C

1 (13 h) 2 (19 h) 3 (22 h) 1 (18 h) 2 (20 h) 3 (22 h) 1 (10 h) 2 (17 h) 3 (23 h)

NTot 4.5 × 105

(2.4 × 105)1.1 × 105

(3.2 × 105)7.8 × 104

(2.1 × 105)1.5 × 105

(5.5 × 104)1.7 × 105

(4.0 × 104)1.1 × 105

(3.4 × 104)5.2 × 105

(2.1 × 105)1.9 × 105

(0.8 × 105)8.7 × 104

(4.3 × 104)Dg 15.4

(61.5)9.56(37.6)

9.20(63.4)

33.2(124)

42.2(147)

47.8(149)

8.36(80)

12.8(102)

5.51(137)

σ 1.55(2.10)

1.40(2.67)

1.46(2.16)

1.59(1.62)

1.52 (1.47) 1.57(1.44)

2.10(2.08)

1.76(1.94)

4.22(1.9)

289S. Byčenkienė et al. / Atmospheric Research 143 (2014) 279–292

during traffic and residential heating event from ~1 to 2 wasfound. The study results for remaining time suggest that theaerosol in the size range of ~100 nm is responsible for the lowBC concentration level. The higher Angstrom exponentsvalues, except during the rush hours, demonstrate thatparticles larger than about 100 nm in diameter cannot beattributed only to the traffic. These particles mainly originatefrom the regional and urban background because thebiomass burning (heating) was potential sources beside thetraffic.

In the previous studies PM10may have been surrogate forthe total number concentration, as the correlation was foundto be strong (Pekkanen et al., 1997). In contrast to thepresent study there was only a weak correlation betweenPNC (N10–20, N20–90) and PM10 on 7 September 2011,while PNC (N100–300) and PM10 were negatively correlated(−0.344). It shows that the high number concentration, butthe low particulate mass is typical of “traffic event day” andthe mass measurements of the particulate matter were notinfluenced by the road dust. The gaseous pollutants werehighly correlated with each other while PNC in different sizeranges did not show a significant correlation with theprimary gaseous pollutants such as CO (r = 0.277) and NOx

(r = 0.220). The results observed for the traffic event(September 7, 2011) are lower in comparison to thosereported by Morawska et al. (1998), where PNC (0.016 to0.7 μm) was reasonably well correlated with CO (r = 0.45)and NOx (r = 0.40) and was influenced by the vehicleexhaust emission. Ketzel et al. (2003) report an r = 0.81 forthe PNC (0.016–0.73 μm) and the NOx concentration inPlabutsch tunnel, Austria. The differences between thecorrelations in the two different environments can princi-pally be attributed to negligible influence of the varying

Fig. 6. Daily variability of N10–20, N20–90 and N100–300 (cm−3)

meteorological parameters. There were some instanceswhere the r was large on 18 September 2011 (panel B), asthe temporal variation for the PNC and NOx was closelycorrelated, while CO did not follow the same trend. Thisindicates that a major proportion of the atmospheric nitrogendioxide was a secondary product of the atmosphericchemistry. The r values of 22 September 2011 (panel C) alsoshowed that the PNC was strongly correlated with NOx andmoderately with the PM10. The low correlation betweenPM10 and PNC (N10–20) showed that it was not possible toobtain information on the particle number from the conven-tional mass-based air pollution parameters during the “trafficevent day”.

4. Conclusions

The obtained size distributions indicated that the aerosolparticle high-polluted events in the urban backgroundenvironment have the same tendency during the trafficrush hours and the residential heating season. The maximumhourly concentration during all the measurement periodswas observed in February 2011 (110,000 cm−3) and theminimum in June 2010 (100 cm−3). The geometrical meandiameters were found to be Dg = 37.6 nm for the nucleationmode, suggesting traffic influence and Dg = 137 nm for theAitken mode (NTot = 4.25 × 104) showing a potential sourceof the smoke trace from the local residential wood burningfor house residential heating. For the NPF event day, the Dg ofultrafine particles was found to be 80 nm, while photochem-ical NPF event provided the lowest average PNC in the 20–90 nm size range (NTot = 1.50 × 105 cm−3) with an averageof Dg = 33.2 nm.

occurred on 7 (A), 18 (B) and 22 (C) September 2011.

290 S. Byčenkienė et al. / Atmospheric Research 143 (2014) 279–292

PCA confirmed that the variations in the PNC diurnalpatterns were mainly due to the traffic and combustionprocesses as it is indicated by the high correlation betweenthe concentrations of NO and NO2 and the particle numberconcentrations of particles between 20 and 300 nm. Arelationship between particles (N10–20, N20–90) and PM10was not recognizable while that of PNC (N100–300) and PM10was negatively correlated (−0.344). It shows that the highnumber concentration, but low particulate mass is typical of“traffic event day” and mass measurements of particulatematter were not strongly influenced by road dust. During thephotochemical NPF event on 18 September 2011 a majorproportion of the atmospheric nitrogen dioxide was recog-nized as a secondary product of atmospheric chemistry.

The PSCF and CWT receptor models were used to identifythe spatial source distributions of PNC. Heavily industrializedareas in Poland, Germany and Czech Republic were identifiedas high PNC source areas. Furthermore, air masses arriving inVilnius are seasonally dependent. For winter, central Europeancountries, including Poland, Germany and Czech Republic, aswell as eastern part of Russia and eastern part of Belarus hadthe highest PSCF values and were thus the most importantlong-range transport sources. In autumn, the air massestransported from eastern part of Belarus was most polluted(11,200 ± 6300 cm−3), while air masses from northerncountries were relatively clean (4100 ± 7700 cm−3). Samedirection air masses in spring led to higher PNC loadings(12,700 ± 10,300 cm−3) in Vilnius, while the air massestransported from eastern part of Belarus were associated withthe lowest concentration of 5500 ± 12,200 cm−3.

References

2008/50/EC. DIRECTIVE 2008/50/EC OF THE EUROPEAN PARLIAMENT ANDOF THE COUNCIL of 21 May 2008 on ambient air quality and cleaner airfor Europe.

Aalto, P., Hameri, K., Paatero, P., Kulmala, M., Forastiere, F., Cattani, G.,Marconi, A., Cyrys, J., VonKlot, S., Zetzsche, K., Peters, A., Bouso, L.,Castano, G., Palacio, J.A., Sunyer, J., Lanki, T., Pekkanen, J., Sjoval, B.,Berglind, N., Bellander, T., Nyberg, F., 2005. Aerosol number concentra-tion measurements in five European cities using TSI-3022 condensationcounter over three-year period during health effects of air pollution onsusceptible subpopulations. J. Air Waste Manage. Assoc. 55 (8),1064–1076.

Allan, J.D., Alfarra, M.R., Bower, K.N., Coe, H., Jayne, J.T., Worsnop, D.R., Aalto,P.P., Kulmala, M., Hyotylainen, T., Cavalli, F., Laaksonen, A., 2006. Sizeand composition measurements of background aerosol and new particlegrowth in a Finnish forest during quest 2 using an aerodyne aerosolmass spectrometer. Atmos. Chem. Phys. 6, 315–327.

Araujo, J.A., Barajas, B., Kleinman, M., Wang, X., Bennett, B.J., Gong, K.W.,Navab, M., Harkema, J., Sioutas, C., Lusis, A.J., Nel, A.E., 2008. Ambientparticulate pollutants in the ultrafine range promote early atheroscle-rosis and systemic oxidative stress. Circ. Res. 102, 589–596.

Ashbaugh, L.L., Malm, W.C., Sadeh, W.Z., 1985. A residence time probabilityanalysis of sulfur concentrations at Grand Canyon National Park. Atmos.Environ. 19, 1263–1270.

Beverland, I.J., Tunes, T., Sozanska, M., Elton, R.A., Agius, R.M., Heal, M.R.,2000. Effect of long-range transport on local PM10 concentrations in theUK. Int. J. Environ. Health Res. 10, 229–238.

Brook, R.D., Rajagopalan, S., Pope III, C.A., Brook, J.R., Bhatnagar, A., Diez-Roux, A.V., Holguin, F., Hong, Y., Luepker, R.V., Mittleman, M.A., Peters,A., Siscovick Jr., D., Smith, S.C., Whitsel, L., Kaufman, J.D., 2010.Particulate matter air pollution and cardiovascular disease. Circulation121, 2331–2378.

Bukowiecki, N., Dommen, J., Prévôt, A.S.H., Richter, R., Weingartner, E.,Baltensperger, U., 2002. A mobile pollutant measurement laboratory—measuring gas phase and aerosol ambient concentrations with highspatial and temporal resolution. Atmos. Environ. 36, 5569–5579.

Byčenkienė, S., Ulevicius, V., Kecorius, S., 2011. Characteristics of blackcarbon aerosol mass concentration over the East Baltic region from two-year measurements. J. Environ. Monit. 13, 1027–1038.

Cheng, M.D., Hopke, P.K., Barrie, L., Rippe, A., Olson, M., Landsberger, S., 1993.Qualitative determination of source regions of aerosol in Canadian HighArctic. Environ. Sci. Technol. 27, 2063–2071.

Cheung, H.C., Morawska, L., Ristovski, Z.D., 2010. Observation of new particleformation in subtropical urban environment. Atmos. Chem. Phys. 11,3823–3833.

Cyrys, J., Pitz, M., Heinrich, J., Wichmann, H.E., Peters, A., 2008. Spatial andtemporal variation of particle number concentration in Augsburg,Germany. Sci. Total Environ. 401, 168–175.

Daigle, C.C., Chalupa, D.C., Gibb, F.R., Morrow, P.E., Oberdorster, G., Utell, M.J.,Frampton, M.W., 2003. Ultrafine particle deposition in humans duringrest and exercise. Inhal. Toxicol. 15, 539–552.

Dal Maso, M., Kulmala, M., Riipinen, I., Wagner, R., Hussein, T., Aalto, P.P.,Lehtinen, E.J., 2005. Formation and growth of fresh atmospheric aerosol:eight years of aerosol size distribution data from SMEAR II, Hyytiälä,Finland. Boreal Env. Res. 10, 323–336.

Dal Maso, M., Sogacheva, L., Aalto, P.P., Riipinen, I., Komppula, M., Tunved, P.,Korhonen, L., Suur-Uski, V., Hirsikko, A., Kurt ́en, T., Kerminen, V.-M.,Lihavainen, H., Viisanen, Y., Hansson, H.-C., Kulmala, M., 2007. Aerosolsize distribution measurements at four Nordic field stations: identifica-tion, analysis and trajectory analysis of new particle formation bursts.Tellus B 59 (3), 350–361.

Draxler, R.R., Rolph, G.D., 2014. HYSPLIT (HYbrid Single-Particle LagrangianIntegrated Trajectory) Model access via NOAA ARL READY Website(http://ready.arl.noaa.gov/HYSPLIT.php) NOAA Air Resources Laborato-ry, Silver Spring, MD.

Einax, J.W., Zwanziger, H.W., Geiss, S., 1997. Chemometrics in EnvironmentalAnalysis. VCH Verlagsgesellschaft mbH, Weinheim 384.

Fahrmeir, L., Hamerle, A., Tutz, G., 1996. Multivariate statistische Verfahren,2nd ed. de Gruyter, Berlin 902 (German).

Fernández-Camacho, R., Rodr'ıguez, S., de la Rosa, J., Snchez de la Campa,A.M., Viana, M., Alastuey, A., Querol, X., 2010. Ultrafine particleformation in the inland sea breeze airflow in Southwest Europe.Atmos. Chem. Phys. 10, 9615–9630.

Gao, J., Wang, J., Cheng, S.H., Xue, L.K., Yan H.Z., Hou, L.J., Jiang, Y.Q., Wang,W.X., 2007. Number concentration and size distributions of submicronparticles in Ji'nan urban area: characteristics in summer and winter.J. Environ. Sci. 19 (12), 1466–1473.

Harris, J.S., Maricq, M.M., 2001. Signature size distributions for diesel andgasoline engine exhaust particulate matter. J. Aerosol Sci. 32, 749–764.

Harrison, R.M., Beddows, D.C.S., Dall'Osto, M., 2011. PMF analysis of wide-range particle size spectra collected on a major highway. Environ. Sci.Technol. 45, 5522–5528.

Holmes, N.S., 2007. A review of particle formation events and growth in theatmosphere in the various environments and discussion of mechanisticimplication. Atmos. Environ. 41, 2183–2201.

Holmes, N.S., Morawska, L., Mengersen, K., Jayaratne, E.R., 2005. Spatialdistribution of submicrometre particles and CO in an urban microscaleenvironment. Atmos. Environ. 39, 3977–3988.

Hopke, P.K., Hwang, I., 2007. Estimation of source apportionment andpotential source locations of PM 2.5 at a west coastal IMPROVE site.Atmos. Environ. 41, 506–518.

Hu, S., Fruin, S., Kozawa, K., Mara, S., Winer, A.M., Paulson, S.E., 2009.Aircraft emission impacts in a neighborhood adjacent to a generalaviation airport in southern California. Environ. Sci. Technol. 43,8039–8045.

Hussein, T., Puutinen, A., Aatlo, P.P., Makela, J.M., Hameri, K., Kulmala, M.,2004. Urban aerosol number size distributions. Atmos. Chem. Phys. 4,391–411.

Jaenicke, R., 1993. Tropospheric aerosols. In: Hobbs, P. (Ed.), Aerosol–Cloud–Climate Interactions. Academic Press, San Diego, CA, pp. 1–31.

Janhäll, S., Jonsson, A.M., Molnàr, P., Svensson, E.A., Hallquist, M., 2004. Sizeresolved traffic emission factors of submicrometer particles. Atmos.Environ. 38, 4331–4340.

Jeong, C.-H., Evans, G.J., Hopke, P.K., Chalupa, D., Utell, M.J., 2006. Influence ofatmospheric dispersion and new particle formation events on ambientparticle number concentration in Rochester, United States, and Toronto,Canada. J. Air Waste Manage. Assoc. 56, 431–443.

Jeong, U., Kim, J., Lee, H., Jung, J., Kim, Y.J., Song, C.H., Koo, J.-H., 2011.Estimation of the contributions of long range transported aerosol in EastAsia to carbonaceous aerosol and PM concentrations in Seoul, Koreausing highly time resolved measurements: a PSCF model approach.J. Environ. Monit. 13, 1905–1918.

Juda-Rezler, K., Reizer, M., Oudinet, J.-P., 2011. Determination and analysis ofPM10 source apportionment during episodes of air pollution in CentralEastern European urban areas: the case of wintertime 2006. Atmos.Environ. 45 (36), 6557–6566.

291S. Byčenkienė et al. / Atmospheric Research 143 (2014) 279–292

Kabashnikov, V.P., Chaikovsky, A.P., Kucsera, T.L., Metelskaya, N.S., 2011.Estimated accuracy of three common trajectory statistical methods.Atmos. Environ. 45, 5425–5430.

Karvosenoja, N., Tainio, M., Kupiainen, K., Tuomisto, J.T., Kukkonen, J.,Johansson, M., 2008. Evaluation of the emissions and uncertainties ofPM2.5 originated from vehicular traffic and domestic wood combustionin Finland. Boreal Environ. Res. 13, 65–74.

Ketzel, M., Wahlin, P., Berkowicz, R., Palmgren, F., 2003. Particle and trace gasemission factors under urban driving conditions in Copenhagen basedon street and roof level observations. Atmos. Environ. 37, 2735–2749.

Ketzel, M., Wåhlin, P., Kristensson, A., Swietlicki, E., Berkowicz, R., Nielsen,O.J., Palmgren, F., 2004. Particle size distribution and particle massmeasurements at urban, near-city and 15 rural level in the Copenhagenarea and Southern Sweden. Atmos. Chem. Phys. 4, 281–292.

Kittelson, D., Watts, W., 2000. Roadway measurements of diesel exhaustaerosols. 19th Annual AAAR Conference, St. Louis, Missouri.

Kittelson, D., Jason, J., Winthrop, W., Qiang, W., Bukowiecki, N., 2000. Dieselaerosol sampling in the atmosphere. SAE Paper No. 2000–01-2122.

Köhler, A., Som, C., Helland, A., Gottschalk, F., 2008. Studying the potentialrelease of carbon nanotubes throughout the application lifecycle. J. CleanProd. 16 (8–9), 927–937.

Kukkonen, J., Pohjola, M., Sokhi, R.S., Luhana, L., Kitwiroon, N., Rantamäki, M.,Berge, E., Odegaard, V., Slørdal, L.H., Denby, B., Finardi, S., 2005. Analysisand evaluation of selected local-scale PM10 air pollution episodes infour European cities: Helsinki, London, Milan and Oslo. Atmos. Environ.39, 2759–2773.

Kulmala, M., Petäjä, T., Mönkkönen, P., Koponen, I.K., Dal Maso, M., Aalto,P.P., Lehtinen, K.E.J., Kerminen, V.-M., 2005. On the growth of nucleationmode particles: source rates of condensable vapor in polluted and cleanenvironments. Atmos. Chem. Phys. 5, 409–416.

Kumar, P., Fennell, P., Britter, R., 2008. Measurements of particles in the5–1000 nm range close to road level in an urban street canyon. Sci. TotalEnviron. 390, 437–447.

Kumar, D., Kumar, G., Singh, P.C.P., 2010. Fast, easy ethanolysis of coconut oilfor biodiesel production assisted by ultrasonication. Ultrason. Sonochem.17, 555–559.

Kumar, P., Robins, A., Vardoulakis, S., Quincey, P., 2011. Technical challengesin tackling regulatory concerns for urban atmospheric nanoparticles.Particuology 9, 566-571.

Laakso, L., Hussein, T., Aarnio, P., Komppula, M., Hiltunen, V., Viisanen, Y.,Kulmala, M., 2003. Diurnal and annual characteristics of particle massand number concentrations in urban, rural and Arctic environments inFinland. Atmos. Environ. 37, 2629–2641.

Lavanchy, V.M.H., Gäggeler, H.W., Schotterer, U., Schwikowski, M.,Baltensperger, U., 1999. Historical record of carbonaceous particleconcentrations from a European high-alpine glacier (Colle Gnifetti,Switzerland). J. Geophys. Res. 104, 21227–21236.

Liu, Na, Ye, Yu., He, Jianjun, Zhao, Suping, 2013. Integrated modeling ofurban-scale pollutant transport: application in a semi-arid urban valley,Northwestern China. Atmos. Pollut. Res. 4 (3), 306–314.

Marconi, A., Cattani, G., Cusano, M., Ferdinandi, M., Inglessis, M., Viviano, G.,Settimo, G., Forastiere, F., 2007. Two-years of fine and ultrafine particlesmeasurements in Rome, Italy. J. Toxicol. Environ. Health A 70 (3–4),213–218.

Maricq, M.M., Podsiadlik, D.H., Chase, R.E., 1998. Gasoline vehicle particlesize distributions: comparison of steady state, FTP and US06 measure-ments. Environ. Sci. Technol. 32, 2033–2042.

Meyer, N.K., 2012. Particulate, black carbon and organic emissions fromsmall-scale residential wood combustion appliances in Switzerland.Biomass Bioenergy 36, 31–42.

Mijić, Zoran, Stojić, Andreja, Perišić, Mirjana, Rajšić, Slavica, Tasić, Mirjana,2012. Statistical Character and Transport Pathways of AtmosphericAerosols in Belgrade, Air Quality — New Perspective. In: Lopez, Gustavo(Ed.), InTech. ISBN: 978-953-51-0674-6. http://dx.doi.org/10.5772/45873.

Morawska, L., Thomas, S., Bofinger, N., Wainwright, D., Neale, D., 1998.Comprehensive characterization of aerosols in a subtropical urbanatmosphere: particle size distribution and correlation with gaseouspollutants. Atmos. Environ. 32 (14–15), 2467–2478.

Morawska, L., Jayaratne, E.R., Mengersen, K., Jamriska, M., Thomas, S., 2002.Differences in airborne particle and gaseous concentrations in urban airbetween weekdays and weekends. Atmos. Environ. 36, 4375–4383.

Morawska, L., Ristovski, Z., Jayaratne, E.R., Keogh, D.U., Ling, X., 2008.Ambient nano and ultrafine particles from motor vehicle emissions:characteristics, ambient processing and implications on human expo-sure. Atmos. Environ. 42, 8113–8138.

Mordas, G., Kulmala, M., Petäjä, T., Aalto, P.P., Matulevičius, V., Grigoraitis, V.,Ulevicius, V., Grauslys, V., Ukkonen, A., Hämeri, K., 2005. Design andperformance characteristics of a condensation particle counter UF-02proto. Boreal Env. Res. 10, 543–552.

Oberdörster, G., Sharp, Z., Elder, A.P., Gelein, R., Kreyling, W., Cox, C., 2004.Translocation of ultrafine particles to the brain. Inhal. Toxicol. 16,437–445.

O'Dowd, C.D., Jimenez, J.L., Bahreini, R., Flagan, R.C., Seinfeld, J.H., Hameri, K.,Pirjola, L., Kulmala, M., Jennings, S.G., Hoffmann, T., 2002. Marine aerosolformation from biogenic iodine emissions. Nature 417, 632–636.

Park, S.S., Kim, Y.J., Cho, S.Y., Kim, S.J., 2007. Characterization of PM2.5aerosols dominated by local pollution and Asian dust observed at anurban site in Korea during aerosol characterization experiments (ACE)—Asia Project. J. Air Waste Manag. Assoc. 57, 434–443.

Park, K., Yeon Park, Ji, Kwak, Ji-Hyun, Cho, Gang Nam, Kim, Jae-Suk, 2008.Seasonal and diurnal variations of ultrafine particle concentration inurban Gwangju, Korea: observation of ultrafine particle events. Atmos.Environ. 42, 788–799.

Pekkanen, J., Timonen, K.L., Ruuskanen, J., Reponen, A., Mirme, A., 1997.Effects of ultrafine and fine particles in urban air on peak expiratory flowamong children with asthmatic symptoms. Environ. Res. 74 (1), 24–33.

Perez, N., Pey, J., Cusack, M., Reche, C., Querol, X., Alastuey, A., Viana, M.,2010. Variability of particle number, black carbon, and PM(10), PM(2.5),and PM(1) levels and speciation: 25 influence of road traffic emissionson urban air quality. Aerosol Sci. Technol. 44, 487–499.

Pey, J., Rodriguez, S., Querol, X., Alastuey, A., Moreno, T., Putaud, J.P., VanDingenen, R., 2008. Variations of urban aerosols in the WesternMediterranean. Atmos. Environ. 42, 9052–9062.

Plauškaitė, K., Ulevičius, V., Špirkauskaitė, N., Byčenkienė, S., Zelinski, T.,Petelski, T., Ponczkowska, A., 2010. Observation of new particlesformation events in the south-eastern Baltic Sea. Oceanologia 52 (1),53–75.

Pope III, C.A., Burnett, R.T., Thun, M.J., Calle, E.E., Krewski, D., Ito, K., Thurston,G.D., 2002. Lung, cancer, cardiopulmonary mortality, and long-termexposure to fine particulate air pollution. JAMA 287 (9), 1132–1141.

Riuttanen, L., Hulkkonen, M., Dal Maso, M., Junninen, H., Kulmala, M., 2013.Trajectory analysis of atmospheric transport of fine particles, SO2, NOx

and O3 to the SMEAR II station in Finland in 1996–2008. Atmos. Chem.Phys. 13, 2153–2164.

Rodriguez, S., Van Dingenen, R., Putaud, J.P., Martins-Dos Santos, S., Roselli,D., 2005. Nucleation and growth of new particles in the ruralatmosphere of Northern Italy — relationship to air quality monitoring.Atmos. Environ. 39 (36), 6734–6746.

Ruuskanen, J., Tuch, Th., Ten Brink, H., Peters, A., Khlystov, A., Mirme, A., Kos,G.P.A., Brunekreef, B., Wichmann, H.E., Buzorius, G., Vallius, M., Kreyling,W., Pekkanen, J., 2001. Concentrations of ultrafine, fine and PM2.5particles in three European cities. Atmos. Environ. 35, 3729–3738.

Salma, I., Borsós, T., Weidinger, T., Aalto, P., Hussein, T., Dal Maso, M.,Kulmala, M., 2011. Production, growth and properties of ultrafineatmospheric aerosol particles in an urban environment. Atmos. Chem.Phys. 11 (3), 1339–1353.

Shi, J.P., Khan, A.A., Harrison, R.M., 1999. Measurements of ultra fine particlesconcentration and size distribution in the urban atmosphere. Sci. TotalEnviron. 235, 51–64.

Simmonds, P., Manning, A., Derwent, R., Ciais, P., Ramonet, M., Kazan, V.,Ryall, D., 2005. A burning question: can recent growth rate anomalies inthe greenhouse gases be attributed to large-scale biomass burningevents? Atmos. Environ. 39, 2513–2517.

Sioutas, C., Zhang, K.M., Wexler, A.S., Niemeier, D.A., Zhu, Y.F., Hinds, W.C.,2005. Evaluation of particle number distribution near roadways. Part III:traffic analysis and on-road size resolved particulate emission factors.Atmos. Environ. 39, 4155–4166.

Sirois, A., Bottenheim, J.W., 1995. Use of backward trajectories to interpretthe 5-year record of PAN and O3 ambient air concentrations atKejimkujik National Park, Nova Scotia. J. Geophys. Res. 100, 2867–2881.

Spracklen, D.V., Carslaw, K.S., Kulmala, M., Kerminen, V.-M., Sihto, S.L.,Riipinen, I., Merikanto, J., Mann, G.W., Chipperfield, M.P., Wiedensohler,A., Birmili, W., Lihavainen, H., 2008. Contribution of particle formation toglobal cloud condensation nuclei concentration. J. Geophys. Res. 35,L06808. http://dx.doi.org/10.1029/2007GL033038.

Stocker, J., Carruthers, D., 2007. Road vehicle non-exhaust particulatematter: Initial air quality model development and application, modeluncertainty analysis and further model improvements. TRL & CERC,CERC Report FM668/R01/06.

Tuch, T.M., Herbarth, O., Franck, U., Peters, A., Wehner, B., Wiedensohler, A.,Heintzenberg, J., 2006. Weak correlation of ultrafine aerosol particleconcentrations b800 nm between two sites within one city. J. Expo. Sci.Environ. Epidemiol. 16 (6), 486–490.

Ulevicius, V., Byčenkienė, S., Remeikis, V., Garbaras, A., Kecorius, S.,Andriejauskienė, J., Jasinevičienė, D., Mocnik, G., 2010. Characterizationof aerosol particle episodes in Lithuania caused by long-range andregional transport. Atmos. Res. 98 (2–4), 190–200.

Venter, A.D., Vakkari, V., Beukes, J.P., van Zyl, P.G., Laakso, H., Mabaso, D.,Tiitta, P., Josipovic, M., Kulmala, M., Pienaar, J.J., Laakso, L., 2012. An air

292 S. Byčenkienė et al. / Atmospheric Research 143 (2014) 279–292

quality assessment in the industrialised western Bushveld IgneousComplex, South Africa. S. Afr. J. Sci. 108 (9/10).

Virkkula, A., Mäkelä, T., Hillamo, R., Yli-Tuomi, T., Hirsikko, A., Hämeri,K., Koponen, I.K., 2007. A simple procedure for correcting loadingeffects of aethalometer data. J. Air Waste Manag. Assoc. 57,1214–1222.

Wahlin, P., Finn, P., Van Dingenen, R., 2001. Experimental studies of ultrafineparticles in streets and the relationship to traffic. Atmos. Environ. 35,63–69.

Wardoyo, A.Y.P., Morawska, L., Ristovski, Z., Jamriska, M., Carr, S.,Johnson, G., 2007. Size distribution of particles emitted fromgrass fires in the Northern Territory, Australia. Atmos. Environ. 41,8609–8619.

Watson, J.G., Chow, J.C., Park, K., Lowenthal, D.H., 2006. Nanoparticle andultrafine particle events at the Fresno supersite. J. Air Waste Manage.Assoc. 56, 417–430.

Weber, S., 2009. Variation of particle number concentrations and noise onthe local urban scale. The Seventh International Conference on UrbanClimate, 29 June–3 July 2009, Yokohama, Japan.

Wehner, B., Uhrner, U., von L¨owis, S., Zallinger, M., Wiedensohler, A., 2009.Aerosol number size distributions within the exhaust plume of a dieseland a gasoline passenger car under on-road conditions and determina-tion of emission factors. Atmos. Eviron. 43, 1235–1245.

Zwozdziak, A., Samek, L., Sowka, I., Furman, L., Skrętowicz, M., 2012. Aerosolpollution from small combustors in a village. ScientificWorldJournal2012, 956401.