Atmospheric Research 95 (2010) 55–64

Contents lists available at ScienceDirect

Atmospheric Research

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

Prediction of sources of metal pollution in rainwater in Istanbul, Turkeyusing factor analysis and long-range transport models

Nihan Uygur a, Ferhat Karaca a, Omar Alagha b,⁎a Department of Environmental Engineering, Fatih University, 34500, Istanbul, Turkeyb Department of Environmental Engineering, King Faisal University, Dammam, KSA

a r t i c l e i n f o

⁎ Corresponding author. Fax: +90 212 86633412.E-mail address: oalagha@gmail.com (O. Alagha).

0169-8095/$ – see front matter © 2009 Elsevier B.V.doi:10.1016/j.atmosres.2009.08.007

a b s t r a c t

Article history:Received 20 April 2009Received in revised form 5 August 2009Accepted 6 August 2009

In this study,major (Na,K,Al, Ca andMg) and trace (Pb,Ni, Fe, Cu, Cr, Co, andV) elements in43 rainsamples were measured and their concentrations were statistically analysed. The samples werecollected duringOctober 2007 andMay 2008 at the campus of FatihUniversity, which is located inIstanbul at 41.09° longitude and 28.61° latitude and is in the Büyükçekmece watershed area.Rainwater samples were collected using a purpose-built wet-only rain sampler designed andconstructed at the Fatih University Department of Environmental Engineering. Major and tracemetals in rainwater were measured using Graphite Furnace Atomic Absorption Spectrometer(GFAAS) and FlameAtomicAbsorption Spectrometer (FAAS) techniques. Statistical analyses showthat the highest concentrations were observed for Al and Fe, which are mainly lithophilicelements. The lowest concentrations were measured for Co. Crustal Enrichment Factor (EFc)calculations clarified that rainwater samples were not enriched withMg, K, Ca, Na or Fe, but weremoderately to extremely enriched with Cr, Co, Ni, Cu, V, and Pb. The latter elements wereattributed to a mix of local and long-range anthropogenic sources. Factor analysis calculationsresulted in five factors with eigenvalues greater than unity. To further elucidate the geographicalcontribution from these factors, a potential source contribution function (PSCF) model wascalculated from the highest quartile of the factor-scored trajectories for the five extracted factors.The PSCF results show that a significant fraction of the measured anthropogenic pollutants wastransported to the sampling area from source regions in Europe, Russia, southern and northernMediterranean countries and industrial zones west of Turkey.

© 2009 Elsevier B.V. All rights reserved.

Keywords:Rainwater compositionHeavy metalGFAASEnrichment factorFactor analysisBackward trajectoryPSCF

1. Introduction

Istanbul is located between the continents of Asia andEurope, separated by the Bosporus Strait. The city is affected byregional and global pressure systems and located along thetransport pathways of air parcels coming primarily fromEurope, the Black Sea and Russia during winter and autumnand from the Mediterranean and Africa during spring andsummer. For this reason, studies focusing on the potentialcontributionof long-range sources towetprecipitation samplescollected in Istanbul have both local and global importance.

All rights reserved.

Studying wet precipitation in Istanbul is of further importancein light of thedecreasing amount and quality of surfacewater inthe city. Surface water reservoirs in Istanbul were nearlyemptied following an extraordinary drought during thesummer of 2008. The chemistry of the precipitation fromrainfall events that followed this drought period is represen-tative of the main factors characterising the chemical compo-sition of Istanbul'swater reserves. Istanbul and its surroundingshave a richfloraandextensive forest tracts that canbedamagedby acid deposition. Finally, Istanbul is home to hundreds ofprecious and uniquehistorical buildings,monuments and otherstructures, to the extent that the city is known as “the timelesscity.” Acidic precipitation deteriorates monuments and histor-ical buildings, and it is thus important to uncover the sources ofpollution affecting precipitation chemistry.

56 N. Uygur et al. / Atmospheric Research 95 (2010) 55–64

The scientific literature contains a limited number of studiesregarding rainwater chemistry in Istanbul (Gulsoy et al., 1999;Okay et al., 2002; Akkoyunlu and Tayanc, 2003; Başak andAlagha, 2004; Saylan et al., 2002; Saylan et al., 2009). To ourknowledge, the first study carried out was by Gulsoy et al.(1999). Between January and October 1999, they collectedurban precipitation samples in three different regions ofIstanbul having different urbanisation characteristics. Majorionswere analysed, and it was concluded that therewas a cleareffect of urban heating on rain chemistry. The pH of thecollected samples ranged from 3.96 to 8.46 with an average of6.15. Only 18.6% of the collected samples were acidic (i.e.,having a pH value less than 5.6, which characterises typicalprecipitation). Most of the acidic events were recorded duringJanuary, a few in early April, and only one in October. Aerosolscontaining calcium and ammonium were the dominant neu-tralising agents.

Okay et al. (2002) collected wet deposition samples fromAugust 1993 to November 1995 in Kaynarca, located in thesoutheast of Istanbul. They measured major cations and anionsand foundanaverage pHvalueof 5.59, lower than that foundbyGulsoy et al. (1999). In addition to evaluating the seasonality oftheir data, they also used statistical methods to determine thechemical sources of the measured ions using enrichmentfactors based oncrustal andmarine sources. Cyclones identifiedfrom 00 GMT surface maps and 500 hPa geopotential heightmaps were used to evaluate the trajectories of air parcelsaffecting the regionduring rain events. Their trajectoryanalysesshowed high sulphate- and nitrate-containing rain eventsassociatedwith cyclonesoriginating in locations fromEurope toIstanbul.

In another study, Akkoyunlu and Tayanc (2003) collectedwet deposition samples (54 wet-only and 21 bulk samples) infour different regions of Istanbul — Topkapı, Bağcılar, MaltepeandGöztepe— between January andMay2001. Theymeasuredmajor cations and anions and found an average pH of 5.26 fortheir wet samples. This value is the lowest average pH evermeasured in Istanbul (Gulsoy et al., 1999; Okay et al., 2002;Akkoyunlu and Tayanc, 2003; Başak and Alagha, 2004; Saylanet al., 2009). Enrichment factors were used to define possibleenrichment sources, and the results from this analysis were inclose agreement with those of previously mentioned studies.

Başak and Alagha (2004) collected 79 bulk precipitationsamples in Büyükçekmece, Istanbul between October 2001and July 2002. This study was the first in Istanbul to con-centrate on both trace elements and major ions. The averagepH was 5.58, similar to that of the rain samples collected inKaynarca (Okay et al., 2002). Atmospheric sources of ions andtrace elements in the Büyükçekmece data were evaluatedusing a variety of statistical methods, including enrichmentfactors and correlation analyses. They concluded that acidneutralisation in the samples was the result of calcium, notammonium, ions.

In a recent study, Akkoyunlu and Tayanc (2008) sequen-tially collected wet precipitation data during four differentrainstorms. Samples were collected between October 2003and February 2004. They concentrated on ionic compositionand variations during sub-events along with the prevailingatmospheric conditions during the storms. They found thatthe first sub-event of each storm had higher pH values thanthe subsequent events due to atmospheric washout.

The most recent study of wet deposition in Istanbulemployed back trajectory analysis of samples collected fromtwo different regions (urban and forest) in Istanbul betweenNovember 1997 and June 1998 (Saylan et al., 2009). Theauthors designed a sequential rain sampler that collectedsamples from sub-events of a rainstorm at 10-minuteintervals. The arithmetic mean of the pH was 6.0. Most ionsin the samples reached Istanbul from the WSW, WNW andNNE directions. This study served as one of the most impor-tant attempts to evaluate the effects of long-range transporton the chemical composition of rain in Istanbul.

Building upon the previouswork, themain goals of our studyare: i) to evaluate the trace element composition of wet-onlyprecipitation samples collected in Büyükçekmece, Istanbul, ii) toemploy statistical tests and models to differentiate possiblesource factors, and iii) to use backward trajectory analysis to-gether with extracted source factors to calculate the potentialcontributions of distant sources to the trace element content ofthe samples.

2. Methods and materials

2.1. Study area

The sampling area is located in the Büyükçekmece Lakewatershed at Fatih University (41.09° N, 28.61° E), a suburbanarea in Istanbul. The lake provides Istanbul with 30% of itspotable water. The maximum depth does not exceed 4 m. Fishsuchasgreymullet, carpandbreamare found in the lake (Başakand Alagha, 2004). The sampling area and location are given inFig. 1.

The rainwater sampling collection campaign began inNovember 2007 and ran until June 2008. A total of 43 sampleswere collected during this period.

2.2. Wet-only rain sampler

Rainwater samples were collected using a purpose-madewet-only rain sampler designed and manufactured at the AirPollution Laboratory, Department of Environmental Engineer-ing, Fatih University. The sampler allows wet-only samples tobe collected during a rain event and can be kept closed duringthe dry periods to eliminate the contribution of dry deposi-tion. The sampler design is shown in Fig. 2.

Dry deposition sampleswere collected from the dry bucketof the sampler, which consists of a High Density Polyethylene(HDPE) container. When there was no rainfall for a one-weekperiod and the sampler's dry bucket was continually opened,the dry samples were obtained by washing the dry bucketwith 100 mL of de-ionised distilled water.

After a rainfall event, samples were collected and immedi-ately transported to the laboratory. Any snow samples collectedin the sampling bucket were brought to the laboratory andallowed to melt at room temperature before measurement orfiltration.

2.3. Volume, pH and conductivity measurement

The volume, pH and specific conductivity of the rain sampleswere determined immediately after collection. First, the volumewas determined by comparing the volume of the sample with

Fig. 1. Study area showing the sampling location.

57N. Uygur et al. / Atmospheric Research 95 (2010) 55–64

that of a graded standard. Next, the pH and conductivity weremeasured in a small portion (15 mL) of the original sampleusing a pre-calibrated pH and conductivitymeter (Eutech Instru-ments®). This portionwas discarded aftermeasurement to elim-inate any possible contamination. Finally, the samples werefiltered via a Nalgene® polycarbonate filtration system using a0.45-µm cellulose acetate filter. These samples were given an IDnumber and stored in Nalgene® bottles at 4 °C until the time ofanalysis.

2.4. Instrumentation and metal analysis

Rainwater compositionwas determined by identifyingmajorand heavy metals using a Varian FS 240 Atomic AbsorptionSpectrometer (AAS).Major elements (Ca, Na, K,Mg and Al)wereanalysed with a Flame Atomic Absorption Spectrometer (FAAS)while trace elements (Pb, Ni, Cr, Cu, Co, Fe, and V)were analysedwith a Graphite Furnace Atomic Absorption Spectrometer(GFAAS). GFAAS is a very sensitive technique for measuringtrace elements in environmental samples with concentrationsat part per billion levels, as is the case for rainwater samples(Alagha, 2000).

2.5. Field and laboratory blanks

Field and laboratory blanks were collected to ensure thatno significant contamination of the rain samples occurredduring the sampling, handling, filtration or measurementsteps. Four field blanks and one laboratory blank were used.The detection limit (DL) for rain was calculated to be threetimes the standard deviation of the blank values. The fieldblank calculations showed that for all themeasured elements,the contribution from contaminants was not significant (lessthan 10% of the total concentration).

2.6. PSCF analysis

In this study, the HYSPLIT (HYbrid Single-Particle La-grangian Integrated Trajectory) model (Draxler and Rolph,2003; Rolph, 2003) was used to calculate back trajectories.The calculated trajectories were subsequently used toanalyse the possible long-range sources contributing to thechemistry of the collected rain samples. Potential sourcecontribution function (PSCF) analysis (Polissar et al., 2001,Koçak et al., 2009; Türküm et al., 2008, Güllü et al., 2005)wasused to define the anthropogenic sources of measured trace

Fig. 2. Wet-only rain sampler.

58 N. Uygur et al. / Atmospheric Research 95 (2010) 55–64

elements in the rain samples. This method facilitates thedevelopment of a probability field that equates highmeasured values at the receptor site with their likely sourcelocations. The PSCF of an element x in sub-region ij is givenby Eq. (1):

PSCFx; j = Wi; j ×

PMi; jPNi; j

" #; ð1Þ

where ∑Ni, j is the total number of trajectory segments inthe ijth sub-region during the entire study period and∑Mi, j

is the total number of “episode” or “polluted” trajectorysegments in the same ijth sub-region during the period.

Since the PSCF is computed as a ratio of the counts ofselected events (Mij) to the counts of all events (Nij), Mij islikely small relative to Nij. Values related to sparse trajectorycoverage of the more distant grid cells may result in a PSCFijwith high uncertainty around an apparent high value(Polissar et al., 2001). There is more statistical stability inthe calculated value for large values of N. Thus, to reduce theeffect of small values of Nij, an arbitrary weight function Wi, j

is included to better reflect the uncertainty in the values forthese cells.Wi, j is the number of end points per cell, calculatedby dividing the PSCF grid by the total number of trajectoryendpoints. The PSCF grid covers the northern hemispherewith 5500 cells at 1°×1° resolution between latitudes of 20°Nand 70°N and longitudes of 50°W and 70°E. The total numberof trajectory endpoints is 138,718. On average, there are

therefore approximately 25 end points per cell. The weightfunction used is given in Eq. (2).

Wi; j =

1:00 for Ni; j > 250:70 for 5 < Ni; j � 250:42 for 2:5 < Ni; j � 50:17 for Ni; j � 2:5

8>><>>:

9>>=>>; ð2Þ

Factor score values obtained from factor analysis of rainsample chemical components were used to select the Mi, j

trajectory segments. Events with high factor loadings (greaterthan 0.5) in individual extracted factors were chosen as theepisodes (the threshold values) for the PSCF analysis.

3. Result and discussion

3.1. Summary statistics

Summary statistics for the measured trace elements areshown in Table 1. Aluminium and iron show normal distribu-tions with low skewness and kurtosis values. This is a generalfeature of lithophilic elements, where high concentrations in soilcannot be easily perturbed by other sources. All other elementsshow high skewness and kurtosis values, which indicateabnormal distribution patterns. The highest observed concentra-tions were those of Al and Fe, which are primarily lithophilicelements. The lowest concentration observedwas that of Co. Theaverage volume-weighted pH (indicative of hydrogen ionconcentration) was 6.9, in the alkaline range, with extreme of

Table 1Summary statistics for the soluble fraction of rainwater.

Al Ca Co Cr Cu Fe K Mg Na Ni V Pb Cond. pH

µg/L µS/cm –

Average 7660 660 0.32 0.58 1450 2750 150 110 1420 0.77 6.63 1.47 24.2 6.96Std. dev. 4750 840 0.67 0.74 2100 3310 330 190 2550 1.01 5.84 2.68 21.5 1.25COV 62% 128% 211% 127% 145% 120% 222% 179% 179% 130% 88% 183% 89% 18%Minimum 780 30 0.01 0.04 30 60 3 4 11 0.05 0.01 0.01 1.00 4.3Maximum 1610 3580 4.26 4.05 1330 1540 1610 1340 1640 6.5 28.7 11.1 109. 9.26Skewness 0.12 5.82 13.3 8.61 11.9 5.79 10.4 16.9 12.6 11.6 6.94 7.84 6.68 −1.36Std. error ofskewness

0.36 0.36 0.36 0.36 0.36 0.36 0.36 0.36 0.36 0.36 0.36 0.36 0.36 0.36

Kurtosis −2.50 4.51 33.6 14.3 27.7 4.99 17.2 51.8 31.8 28.2 9.00 8.60 7.44 −0.97Std. error ofkurtosis

0.71 0.71 0.71 0.71 0.71 0.71 0.71 0.71 0.71 0.71 0.71 0.71 0.71 0.71

59N. Uygur et al. / Atmospheric Research 95 (2010) 55–64

4.3 and 9.2. The rainwater was thus not acidic on average butsome possible acidic events occurred.

Our samples were less acidic (higher pH) than generalrainwater (pH slightly below 6), in agreement with the scientificliterature that the average pH of precipitation in Turkey slightlylower or higher than 6 (Gulsoy et al., 1999; Tuncer et al., 2001;Okay et al., 2002; Akkoyunlu and Tayanc, 2003; Saylan et al.,2009). Our average value is higher than reported in mostprevious studies, including the average volume-weighted pH of5.1 observedbyBaşak andAlagha (2004), but similar values havealso been reported (Demirak et al., 2006; Demirak, 2007;Theodosi et al., in press). The higher pH found here is mostlydue to high loading of one ormore of the alkaline species, such asCaCO3 and NH3. Our study region is located in a suburban part ofIstanbul, and there is heavy agricultural activity near thesampling station. In addition, most of the samples with high pHwere collected during spring (March, April and May), whenammonium-based fertilisers were used intensively in the region.CaCO3 is another important alkaline species and originates fromcontinental sources (natural). Previous studies reported thataerosols collected in several regions of Turkey had been enrichedby Ca-containing particles generated by wind erosion (Güllüet al., 2005; Koçak et al., 2009). In this study, rainwater sampleswith high pH values mostly contain relatively higher concentra-tions of Ca ion than the rest of the samples. In addition, Theodosiet al. (in press) reported that Saharan dust events can increasethe pH of rainwater to values as high as 8. The spring monthsstudied here (March, April and May) are the typical monthsduring which long-range transport of Saharan dust is observedover the region (Theodosi et al., in press). In the following

Fig. 3. Crustal Enrichment Factors (EFc) for elements measured in rainwater.

analysis, the effects of desert areas (Saharan and Arabian) on thefactor groups are given in Figs. 6–8.

3.2. Crustal Enrichment Factor (EFc)

The Crustal Enrichment Factor (EFc) provides a simple anduseful way to scale elements according to their degree ofperturbation relative to soil-originated elements. The EF analysisused here follows the procedure employed by Başak and Alagha(2004). Briefly, the EFc was calculated according to Eq. (3):

EFc =Cx =CAlð ÞRainSample

Cx =CAlð Þsoil; ð3Þ

where (Cx)RainSample is the concentration of the element in therain water sample, (Cx)soil is the concentration of the sameelement in the soil, and CAl is the concentration of the referenceelement (here, Al) in the same sample and soil.

Resultant values of EFc are given in Fig. 3. As expected, Mg, K,Ca,Na andFearenot enriched in the collected rainwater samples.On theotherhand, theheavyelementsCr, Co,Ni, Cu,V, andPbaremoderately to extremely enriched relative to the referencevalues. The latter elements originate primarily from a variety ofanthropogenic sources. Thedegreeof enrichmentdependson thetype, proximity and extent of individual sources. The possiblesources of these elements will be discussed in the followingsection.

Table 2Factor loading matrix after Varimax rotation.

Factor 1 Factor 2 Factor 3 Factor 4 Factor 5

Al 0.61Ca 0.60Co 0.78CrCu 0.72Fe 0.82K 0.71Mg 0.93Na 0.89Ni 0.70Pb 0.71 0.54V 0.86

Insignificant factor loadings (<0.5) are not shown in the table.

Table 3Eigenvalues and percentage variance for factors derived by factor analysis.

Factornumber

Eigenvalue Percent ofvariance

Cumulativepercentage

1 2.15 17.96 17.962 1.97 16.39 34.353 1.42 11.87 46.224 1.24 10.37 56.595 1.09 9.12 65.716 0.97 8.08 73.797 0.79 6.57 80.378 0.78 6.46 86.829 0.68 5.68 92.5010 0.48 4.01 96.5211 0.27 2.25 98.7712 0.15 1.23 100.00

60 N. Uygur et al. / Atmospheric Research 95 (2010) 55–64

3.3. Factor analysis (FA)

To statistically bin the measured pollutants into groups ofsimilar variance, Factor Analysis (FA) was used. Applying FAresulted in five factors with eigenvalues greater than unity. The

Fig. 4. Trajectory ensemble for the rain even

first factor (Factor 1) is high in Ca, Co, andMg and explains 17.96%of the variance. Magnesium and Ca are typically crustal elements,whileCo sources aregenerally anthropogenic.AtmosphericCocanbeattributed to coal combustion,mining, ore smelting facilities, airtraffic, road traffic (especially from highways), or specificindustries such as copper-nickel smelting, processing of Co-containing ores and production and use of cobalt chemicals. Thus,this factor represents a mix of crustal and anthropogenicsignatures, the latter likely due to local industrial activities(especially coating and casting) in the nearby area. The EFcanalysis in Fig. 2 further supports this conclusion: bothMg and Caare unenriched, while Co is highly enriched.

The second factor (Factor 2) includes Cu, Fe and Pb andexplains 16.39% of the total variance. Factor 2 also showssubstantial mixing, with Fe unenriched (Fig. 2) and Cu and Pbhighly enriched by anthropogenic activities such as coatingand combustion. The strong correlation between these ele-ments could be attributed to road traffic, as many researchershave suggested (Morselli et al., 2003). It should, however, benoted that leaded gasolinewas banned in Turkey at the end of2005, so the correlated contribution of Cu and Pb in Factor 2

t observed on 07.11.2007 at 08:00 am.

Fig. 5. PSCF plot for Factor 1 group elements in rain samples.

61N. Uygur et al. / Atmospheric Research 95 (2010) 55–64

may result from long-range transport of traffic emissionsrather than local vehicle sources. Alternatively, this factormay be explained by re-suspension of Cu- and Pb-enrichedparticles that had previously settled.

Factor 3 includes K and Ni. This appears to be the result ofwood combustion mixed with crustal sources. The fourthfactor includes Na and Pb. Na is not enriched, while Pb isextremely enriched and originates mainly from anthropo-genic combustion and incineration activities. Finally, Factor 5includes Al and V and also indicates mixing, with Al having acrustal origin and V an anthropogenic origin (Table 2).

Fig. 6. PSCF plot for Factor 2 group

The eigenvalues and percentage variance explained byeach factor are given in Table 3. The selected five factorstogether account for 65.7% of the total variability in theoriginal data.

3.4. Backward trajectory analysis

Backward trajectories computed using the HYSPLIT modelwere used to evaluate the contribution of long-range sourcesto measured rain chemistry. The trajectories started from analtitude of 1500 m above average ground level, which is

elements in rain samples.

Fig. 7. PSCF plot for Factor 3 group elements in rain samples.

62 N. Uygur et al. / Atmospheric Research 95 (2010) 55–64

generally accepted as the altitude of transport of rain clouds(Başak and Alagha, 2004). Back trajectories computed from areceptor site are only accurate to within 20% of the distancetravelled (on average), and any individual trajectory may becompletely incorrect (Owega et al., 2006). Here, the uncer-tainty may be greater due to the limited number of rainsamples (43). To increase the robustness of the trajectoryanalysis, we used the trajectory ensemble option provided byHYSPLIT. The trajectory ensemble option calculates multipletrajectories from the receptor site. Each member of thetrajectory ensemble is calculated by offsetting the meteorol-ogical data by a fixed grid factor (one meteorological gridpoint in the horizontal and 0.01 sigma units in the vertical).

Fig. 8. PSCF plot for Factor 4 grou

This results in 27 members for all possible offsets in X, Y, andZ. A sample trajectory ensemble is given in Fig. 4.

The meteorological systems influencing the rainwatersamples collectedhere came from theeastern and southeasternregions of Europe. Back trajectories show that air parcelsarriving in Istanbul were primarily transported along four aircorridors. Most (about 35%) of the systems were transportedfrom the northeast over the Black Sea region. An additional 10%were transported over Italy and Greece. A further 25% camefrom southern regions, especially from African countries in theSaharan region, and travelled over the southernMediterraneancoast. The remaining 25% originated from the northeast andpassed over Poland, Ukraine, Germany and England.

p elements in rain samples.

Fig. 9. PSCF plot for Factor 5 group elements in rain samples.

63N. Uygur et al. / Atmospheric Research 95 (2010) 55–64

3.5. Source-oriented modelling with PSCF

Figs. 5–9 showthePSCFof the factor scores resulting fromthefactor analysis calculations. The first factor was loaded withanthropogenic heavy metals (Co and Cr) in addition to Ca andMg. The PSCF based on Factor 1 is given in Fig. 5 and shows thatScandinavian countries, Russia, and Ukraine contributed signif-icantly to the observedmetal pollution. The PSCF based on Factor2 is given in Fig. 6. This figure demonstrates that the observedhigh concentrations of metals (Cu, Fe and Pb) came primarilyfrom Russia and Middle Eastern countries (Jordan, Saudi Arabiaand Iraq), with moderate influence from local sources. The PSCFof Factor 3 is given in Fig. 7. The episodic observations of highlevels of K and Ni come mainly from Russia and with someinfluence from Finland and moderate contributions from localsources. Fig. 8 shows the contribution to observed Na and Pbpollution of both local sources and long-range transport from thenorth and the south. Sodium is typically a lithophilic andmarineelement that is entrained by air masses from the south passingover desert areas and theMediterranean Sea. On the other hand,the high Pb is attributable to air masses from the north travellingover the developed Scandinavian countries and Russia. The finalfactor showedhigh concentrations of Al andV. Its PSCF, plotted inFig. 9, demonstrates that Western Europe and Russia weresignificant sources of observed heavy metal pollution; however,local contributions from western Turkey were not negligible.Generally, the observed episodic events were mainly influencedby long-range transport. They were impacted to lesser extent bylocal pollution sources, especially the industrialised areas locatedin the western part of Turkey.

4. Conclusion

In this study, air quality was evaluated by measuringrainwater chemistry near a suburban area of Istanbul, thelargest city in Turkey. Rainwater sampleswere collected using apurpose-built wet-only rain sampler.

The measured parameters were assessed statistically todetermine common trends and possible source regions. Theaverage volume-weighted pHwas 6.9, with extreme values of4.3 and 9.2. The sampled rainwater was therefore generallynot acidic.

Source-oriented models were used to elucidate thepossible local and remote sources of the observed episodicpollution events. An enrichment factor was calculated thatrevealed that the major elements Mg, K, Ca, Na and Fe werenot enriched in the collected rainwater samples while thetrace elements Cr, Co, Ni, Cu, V, and Pb were moderately toextremely enriched. This finding indicates a strong anthro-pogenic contribution to the observed values.

To identify the sources of these pollutants, factor analysiswas performed. The results revealed five factors with mixedanthropogenic and crustal origins. The factor scores of thesefive factors were inspected and the trajectories correspondingto these episodic values were calculated. The source regionsfor the polluted rain events measured here are depicted onthe relevant maps.

These calculations reveal that Western Europe and Russiawere significant sources of heavymetal pollution observed at thecollection site. Local contributions from the western part ofTurkey were not negligible. Generally, the observed episodicevents were mainly influenced by long-range transport, while alesser role was played by local sources, especially those in theindustrialised areas of western Turkey.

Acknowledgments

This work was supported by the Fatih University ScientificResearch Projects Fund, project number P50080801-1. Theauthors gratefully acknowledge the NOAA Air Resources Labo-ratory (ARL) for the provision of the HYSPLIT transport anddispersion model and the READY website (http://www.arl.noaa.gov/ready.html) used in this publication.

64 N. Uygur et al. / Atmospheric Research 95 (2010) 55–64

References

Akkoyunlu, B.O., Tayanc, M., 2003. Analyses of wet and bulk deposition in fourdifferent regions of Istanbul, Turkey. Atmospheric Environment 37 (25),3571–3579.

Akkoyunlu, B.O., Tayanc, M., 2008. Four storms with sub-events: samplingand analysis. Environment International 34 (5), 606–612.

Alagha, O., 2000. Deposition Fluxes of Pollutants Over Black Sea ForestRegion, Ph.D Thesis, Middle East technical University, Ankara.

Başak, B., Alagha, O., 2004. The chemical composition of rainwater overBüyükcekmece Lake, Istanbul. Atmospheric Research 71, 275–288.

Demirak, A., 2007. The influence of a coal-fired power plant in Turkey on thechemical composition of rain water in a certain region. EnvironmentalMonitoring and Assessment 129, 189–196.

Demirak, A., Balcı, A., Karaoğlu, H., Tosmur, B., 2006. Chemical characteristicsof rainwater at an urban site of South Western Turkey. EnvironmentalMonitoring and Assessment 123, 271–283.

Draxler, R.R., Rolph, G.D., 2003. HYSPLIT (HYbrid Single-Particle LagrangianIntegrated Trajectory) Model. NOAA Air Resources Laboratory, SilverSpring, MD. access via NOAA ARL READY Website (http://www.arl.noaa.gov/ready/hysplit4.html).

Güllü, G., Doğan, G., Tuncel, G., 2005. Atmospheric trace element and majorion concentrations over the eastern Mediterranean Sea: identification ofanthropogenic source regions. Atmospheric Environment 39 (34),6376–6387.

Gulsoy, G., Tayanc, M., Erturk, F., 1999. Chemical analyses of the major ions inthe precipitation of Istanbul, Turkey. Environmental Pollution 105 (2),273–280.

Koçak, M., Mihalopoulos, N., Kubilay, N., 2009. Origin and source regions ofPM10 in the Eastern Mediterranean atmosphere. Atmospheric Research92 (4), 464–474.

Morselli, L., Olivieri, P., Brusori, B., Passarini, F., 2003. Soluble and insolublefractions of heavy metals in wet and dry atmospheric depositions inBologna, Italy. Environmental Pollution 124, 457–469.

Okay, C., Akkoyunlu, B.O., Tayanc, M., 2002. Composition of wet deposition inKaynarca, Turkey. Environmental Pollution 118 (3), 401–410.

Owega, S., Khan, B.U.Z., Evans, G.J., Jervis, R.E., Fila, M., 2006. Identification oflong-range aerosol transport patterns to Toronto via classification ofback trajectories by cluster analysis and neural network techniques.Chemometrics and Intelligent Laboratory Systems 83 (1–7), 26–33.

Polissar, A.V., Hopke, P.K., Harris, J.M., 2001. Source regions for atmosphericaerosol measured at Barrow, Alaska. Environmental Science & Technol-ogy 35 (21), 4214–4226.

Rolph, G.D., 2003. Real-time Environmental Applications and Display sYstem(READY) Website. NOAA Air Resources Laboratory, Silver Spring, MD.(http://www.arl.noaa.gov/ready/hysplit4.html).

Saylan, L., Toros, H., Sen, O., 2002. Temporal variability of the precipitationchemistry in the forest area of Istanbul, Turkey. International Journal ofEnvironment and Pollution 18 (5), 508–516.

Saylan, L., Toros, H., Sen, O., 2009. Back trajectory analysis of precipitationchemistry in the urban and forest areas of Istanbul, Turkey. Clean-Soil AirWater 37 (2), 132–135.

Theodosi, C., Markaki, Z., Mihalopoulos, M., in press. Iron speciation,solubility and temporal variability inwet and dry deposition in the EasternMediterranean, Marine Chemistry (Available online 9 July 2008).doi:10.1016/j.marchem.2008.05.004.

Tuncer, B., Bayar, B., Yesilyurt, C., Tuncel, G., 2001. Ionic composition ofprecipitation at the central Anatolia (Turkey). Atmospheric Environment35, 5989–6002.

Türküm, A., Pekey, B., Pekey, H., Tuncel, G., 2008. Comparison of sourcesaffecting chemical compositions of aerosol and rainwater at differentlocations in Turkey. Atmospheric Research 89 (4), 306–314.