Partitioning characteristics of PCBs in urban surface films

This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institution

and sharing with colleagues.

Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party

websites are prohibited.

In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information

regarding Elsevier’s archiving and manuscript policies areencouraged to visit:

http://www.elsevier.com/copyright

http://www.elsevier.com/copyright

Author's personal copyARTICLE IN PRESS

Partitioning characteristics of PCBs in urban surface films

Rosa W. Wu a,b, Tom Harner a,�, Miriam L. Diamond b, Bryony Wilford a,1

a Science and Technology Branch, Environment Canada, 4905 Dufferin Street, Toronto, Ont., Canada M3H 5T4b Department of Geography, University of Toronto, 100 St. George Street, Toronto, Ont., Canada M5S 3G3

a r t i c l e i n f o

Article history:

Received 20 November 2007

Received in revised form

13 February 2008

Accepted 1 March 2008

Keywords:

Surface film

PCBs

Partitioning

a b s t r a c t

Urban areas are characterized by impervious surfaces on which a thin surface film layer

forms. The film is an environmental compartment whereby semi-volatile organic

compounds partition between the organic phase of the film and the ambient air. This

study investigates the partitioning of polychlorinated biphenyls (PCBs) between surface

films and ambient air. The film–air partition coefficient (Kfa) was calculated through

laboratory experiments using ambient film samples, and then plotted against the log of

the octanol–air partition coefficient (Koa). The resultant log–log relationship is analogous

to that between Koa and Kp (gas–particle partition coefficient) where the latter has been

measured for many compound classes, and thus, Kfa can be used in models that consider

the impact of the surface film on chemical dynamics. Ethylene vinyl acetate (EVA) was

investigated as a surrogate film and EVA–air partition coefficients (KEVA–air) values were

measured. The relationship between log Kfa and log KEVA–air is stronger than that between

log Kfa and log Koa, which leads to the preliminary conclusion that EVA may be a more

suitable surrogate for ambient surface films than octanol. Film–air partition quotients

(Qfa) were derived based on field measurements of PCBs in the ambient air and

concentrations in surface films. Qfa depends not only on the characteristics of the organic

phase, but also on external factors, including but not limited to, temperature and ambient

air concentrations. Finally, differences in PCB partitioning between film collected at a

rural site versus an urban site highlights the importance of the composition of the organic

bulk phase in determining chemical sorption into and subsequent volatilization out of the

surface films.

& 2008 Elsevier Ltd. All rights reserved.

1. Introduction

Urban areas are dominated by impervious surfaces, onwhich an atmospherically derived surface film layer forms(Diamond et al., 2000). Gingrich et al. (2001) postulatedthat the mechanisms driving film formation is thecondensation and deposition of gas-phase and particle-phase organic compounds. The presence of the film isimportant in the dynamics of chemical movement and

partitioning in urban environments because it can act as asink or a source for organic compounds to and from theambient air. Deposition of particles plays an importantrole in the development of this surface film (Wu et al.,2008; Priemer and Diamond, 2002; Law and Diamond,1998). Finizio et al. (1997) suggested that organic matteron aerosols behaves similarly to octanol in that solutes areabsorbed into this portion of the aerosol particle. Previousstudies (Wu et al., 2008) have shown that this surfacematerial is constantly being washed off and re-generated,providing a dynamic compartment into which compoundscan partition and a direct link between surface air andreceiving water. Film accumulation is greater in an urbanthan rural area, likely due to a greater number of chemicalsources, higher atmospheric concentrations of numerous

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/atmosenv

Atmospheric Environment

1352-2310/$ - see front matter & 2008 Elsevier Ltd. All rights reserved.

doi:10.1016/j.atmosenv.2008.03.009

� Corresponding author. Tel.: +1416 739 4837.

E-mail address: [email protected] (T. Harner).1 Current address: Centre for Environment, Fisheries and Aquacul-

tural Science, Remembrance Ave., Burnham-on-Crouch, Essex, CM0 8HA,

UK.

Atmospheric Environment 42 (2008) 5696– 5705


compounds, and a higher deposition rate of particlesin urban than rural locations (Gingrich et al., 2001; Wuet al., 2008).

Numerous chemicals have been identified in surfacefilms, including trace contaminants (e.g. PAH, PCBs,PBDEs), biogenic compounds (e.g. alkanoic acids) andinorganics (Liu et al., 2003; Butt et al., 2004; Lam et al.,2005). In urban locations, the surface film can act as animportant buffer or transient sink in removing PCBs fromambient air (Wu et al., 2008). In that context, Wu et al.(2007) estimated that based on an average estimated 44%horizontal impervious surfaces in downtown Toronto,Canada, which spans approximately 641 km2 in total area(Boyd et al., 1993; City of Toronto, 2005) and incorporatinga surface index of 2 that was applied in Priemer andDiamond (2002) to calculate a three-dimensional surfacevolume, a film of approximately 70 nm (�1 month of filmaccumulation) can retain up to 56–226 g of PCBs at anygiven time. The mass of PCBs removed by the film isactually greater than this if we consider the continuouscycle of film wash-off and regeneration. Because thesurface films play a role in urban fate processes, under-standing the partitioning behavior of chemicals betweenair and surface films can help improve descriptions ofthe processes in urban contaminant fate models(Diamond et al., 2001).

This study follows previous work by Wu et al. (2008)that showed that urban surface films developed at a linearrate, averaging between 1.6 and 2.7 nm day�1. Surface filmgrowth is determined by temperature, meteorology andby local sources of particles, both natural and anthropo-genic. This second study builds on the former work andexplores the partitioning behavior of surface films usinglaboratory and field approaches. First, film–air partition-ing was examined using a synthetic film of ethylene vinylacetate (EVA; Elvax 40 W, DuPont, Canada). Wilcocksonand Gobas (2001) as well as Harner et al. (2003) have usedthis synthetic film to approximate the uptake of organiccontaminants in lipid-like compartments found in thenatural environment; compartments which include or-ganic carbon in soil, vegetation and aerosols. Here,experiments with EVA were undertaken to criticallyevaluate using EVA as a surrogate for surface films.Second, we derived film–air quotients and coefficientsfrom actual surface films collected in the field. While onlyPCBs were examined in the present study, the experi-mental procedure has the potential to be extended toother compound classes of interest, especially emergingchemicals whose emissions and impacts are not yetwell studied.

2. Materials and methods

2.1. Ethylene vinyl acetate synthetic film

2.1.1. Film preparation

EVA was used as a surrogate film to validate themethod for measuring film–air partition coefficients. EVAwas chosen because of its relatively high capacity forretaining POPs, and previous experiments have already

investigated EVA–air partitioning, showing it is a suitablealternative for equilibrium sampling (Wilcockson andGobas, 2001; Harner et al., 2003). An EVA stock solutionwas prepared by dissolving 16 g of EVA pellets in 200 mLof DCM, then stored in an amber-colored glass jar at �4 1C.The EVA solution was then spiked with 0.5 mL of a100 pgmL�1 standard mix of 10 PCBs (PCB-3, PCB-15, PCB-18, PCB-29, PCB-28, PCB-53, PCB-52, PCB-49, PCB-61, andPCB-66) and organochlorine pesticides (OCPs—a-HCH, b-HCH, lindane (d-HCH), heptachlor, aldrin, endosulfan I,and endrin). PCB standards were obtained from AccuS-tandard & Ultra Scientific and OCP standards from US EPA.

Approximately 50 g of clean 3-mm diameter Pyrexglass beads were coated with a single coating of thespiked EVA solution. The glass beads were purchased fromFisher Scientific and prepared by cleaning with multiplerinses (n ¼ minimum 4) of dichloromethane (DCM) thenbaking in a muffle furnace for �12 h at 400 1C. This wasdone by filling a solvent-rinsed jar with clean, dry glassbeads then shaking the beads in �25 mL of the spiked EVAsolution (or enough EVA solution to cover all the beads)until all the beads were well coated (where one rinsinggives a film coating of approximately 1 mm in thickness).Excess solution was decanted off, and the beads pouredinto a stainless steel mixing bowl and swirled forapproximately 1 min to prevent the beads from stickingtogether and to allow the DCM to evaporate.

2.1.2. Film mass and thickness

In order to calculate an EVA–air partition coefficient(KEVA–air), the concentration of the compounds in the EVA(CEVA) is needed, based on the volume of EVA coated onthe beads, and which is calculated with the EVA mass anddensity. To determine the mass and thickness of the EVAfilm, EVA-coated beads were rinsed with two washes of25 mL of DCM to extract the spiked EVA solution. Thesolution was decanted into a pre-weighed glass vessel andthe residual DCM on the glass beads was allowed toevaporate so that only EVA remained. EVA mass wasdetermined gravimetrically, and a film thickness andvolume calculated based on bead surface area and EVAdensity of 965 kg m�3. Once the compound mass in thefilm is determined, a concentration in EVA in mg L�1 canbe calculated.

2.1.3. EVA–air partition coefficient (KEVA–air)

Once EVA mass was determined, to obtain theconcentration of the spiked chemicals in the EVA film,5 mL of methanol was added to the EVA, as EVA is notsoluble in methanol, and the mixture sonicated for 1 h toallow for dialysis of the spiked chemicals into themethanol. The methanol was decanted from the mixtureand the extraction was repeated for a total of 10 mL ofmethanol. The methanol was then evaporated with asteady stream of nitrogen to about 0.5 mL, solventexchanged into ethyl acetate and finally into iso-octanefor analysis. A chemical concentration in EVA wascalculated (CEVA) as the mass of the recovered PCBs andOCPs divided by the volume of EVA, in units of pg L�1.Next, the amount of chemical that would partition intothe air phase, or the air concentration, was measured to

R.W. Wu et al. / Atmospheric Environment 42 (2008) 5696–5705 5697


calculate the KEVA–air partition coefficient. To find the airconcentration (Cair), approximately 20 g of spiked EVA-coated beads were placed in a small glass column, then astream of nitrogen was blown through the EVA-coatedbeads at approximately 80–100 mL min�1. The apparatuswas kept insulated in a water bath at a constanttemperature of 20 1C. As previously demonstrated indeterminations of the octanol–air partition coefficientsusing a similar apparatus (Harner and Bidleman, 1996),chemicals partition out of the organic layer on the beadsand attain their equilibrium concentration in the gasstream. These gas-phase PCBs and OCPs in the glasschamber were collected on a C18 bonded silica trap. TheC18 trap was extracted with 30 mL of a 50:50 DCM:hexanesolution, which was then blown down to 1 mL withnitrogen and solvent exchanged into iso-ocatane foranalysis. Analysis was by gas chromatography–electroncapture detection (GC–ECD), with the temperature pro-gram as outlined in Harner and Bidleman (1996), andusing Mirex as the internal standard to correct for volume.The compounds were identified based on retention timeand quantified against a five-point calibration curve usinga 56-compound mixture for PCBs and a mixture of OCPs,containing the target analytes. An EVA–air partitioningcoefficient was calculated as

KEVA2air ¼CEVA

Cair, (1)

where CEVA is the concentration in EVA in pg L�1 and Cair

the concentration in the air stream, also in pg L�1.The temperature dependency of the KEVA–air value was

also measured by conducting replicate measurements ofthe partition coefficient at 15, 25, 35, and 45 1C byinsulating the column in a water bath.

2.1.4. Quality control/quality assurance

The efficiency of the EVA extraction by DCM was testedby rinsing the experimental glass beads (which werepreviously EVA-coated and already extracted twice withDCM) another time and analyzing the additional rinse fortarget analytes. PCB and OCP recovery samples were taken

by coating the beads with various concentrations of thespiked EVA solution and analyzing by gas chromatogra-phy. Experimental blank samples were also taken bycoating glass beads with non-spiked EVA solution andanalyzing for PCBs and OCPs. When determining theEVA–air partition coefficient values, solvent blanks weretaken, consisting of analyzing the DCM-hexane mixturefor target analytes. C18 bonded silica blanks were alsotaken by washing clean C18 samples with solvent, andanalyzing the solvent for PCBs and OCPs.

The method detection limit (MDL) was taken as theaverage of the C18 and solvent blanks. The MDL values forthe KEVA–air experiments ranged between non-detectedand up to 0.8 ng (for one of the most volatile compounds,PCB-3). The majority of the values were well below 100 ng.There was no pattern as to which compound had lowerMDL values.

2.2. Field-collected surface films

2.2.1. Field sampling

Films were collected at an urban site and at a rural siteunder a variety of scenarios (Table 1) to investigatevarious influences on film and partitioning characteristicsincluding temperature and longer seasonal effects. Detailson the field collection of surface films and the analysis forfilm mass and chemical content can be found in Wu et al.(2008). Samples of surface film were obtained using 3-mm diameter glass beads as a surrogate urban impervioussurface. Glass beads, which were chosen to maximizesurface area and hence, film mass collected on the im-pervious surface, were placed on aluminum mesh trays toallow exposure from all directions. The trays were coveredto protect from precipitation. Prior to field deployment,the glass beads were cleaned by multiple rinses with DCMand baked at 400 1C overnight. Each sample consisted of asingle tray of beads (approximately 75–80 g correspond-ing to 2250–2400 individual 3-mm diameter beads).Thirty bead trays were deployed simultaneously for eachurban study.

Table 1Summary of the sampling studies where real surface organic film was collected for laboratory analysis of film evolution rates and partitioning

experiments

Study Location Sampling period Sampling duration Meteorological conditions Sampling interval

PGB-1 Downtown Toronto,

Physical Geography

Building

September– December

2003

89 days Average daily temperature range

of 2.7– 21.4 1C

Every 1– 2 days

PGB-2 Same July 2004– April 2005 273 days Average daily temperature range

of �9.4 to 22.5 1C

Every 1– 2 weeks

PGB-3 Same Winter (November

2004– January 2005)

Winter—86 days 1 sample collection at

end of sampling period

Spring (March– June

2005)

Spring—77 days

Dorset Dorset, Ontario Winter (November

2004– January 2005)

Winter—95 days 1 sample collection at

end of sampling period

Spring (March– June

2005)

Spring—71 days

R.W. Wu et al. / Atmospheric Environment 42 (2008) 5696–57055698


2.2.2. Film–air partitioning (Qfa and Kfa)

The concentration of PCB in the surface films wasdetermined by the method as described in Wu et al.(2008), where the surface film was collected on 3-mmdiameter glass beads, then removed with two rinses ofdichloromethane. Film mass was determined gravimetri-cally, then the film cleaned using a silica column, andthe extract analyzed by GC–ECD. A film density of1000 kg m�3 was assigned to the surface films, and 15PCBs (representing congeners that are dominant in airacross a range of homolog groups) were analyzed—PCB-8,PCB-18, PCB-28, PCB-52, PCB-49, PCB-101, PCB-118, PCB-153, PCB-105, PCB-137/138, PCB-156/171, PCB-180, andPCB-209, based on retention time comparison with astandard, using a five-point calibration curve. The filmdensity was assumed based on the contributions ofmineral matter (density41000) and organic matter(densityo1000), and assuming a uniform film coatingaround all beads. As Lam et al. (2005) showed that thefilm comprises 95% inorganic constituents, the assumeddensity of 1 g cm�3 is likely an underestimate.

Two different partition values were calculated usingthe ambient film data—the partition quotient (Qfa) andpartition coefficient (Kfa). Qfa values for these ambientfilms were calculated as the ratio of PCB congenerconcentrations in film (Cfilm) and in ambient air (Cair)(obtained from Environment Canada’s Integrated Atmo-spheric Deposition Network (IADN), measured at the Gagebuilding of the University of Toronto, a building locatedapproximately 200 m south of PGB), with Eq. (1) asdescribed for the synthetic EVA films (KEVA–air). Thesevalues are referred to as quotients, rather than coefficientsbecause Cfilm and Cair are continuously fluctuating and notnecessarily in equilibrium. Partition coefficients (Kfa) werederived at room temperature (�22 1C) using the apparatusas discussed in Harner and Bidleman (1996) who weremeasuring octanol–air partition coefficients Koa (Fig. 1),and have shown that equilibrium between octanol and airis established. It is important to note that Qfa as presented

here relates to the bulk film, which contains both organicmaterial and mineral matter. As PCBs and other POPs areexpected to be mainly associated with the organic matterfraction, it would be preferable to calculate a Qfa,OM value,which reflects the organic matter content, whereQfa,OM ¼ Qfa/fOM. However, the organic matter contentdata were not available for these calculations.

2.2.3. Quality control/quality assurance

To correct for errors or contamination that may haveoccurred during sample collection, field blank samples ofglass beads were taken. These consisted of glass beadsthat were treated in the same manner as the samples butwere collected immediately after they were deployed.Solvent blank samples (method blanks) were also takenby analyzing the solvent only for chemical content.

For the Kfa experiments involving ambient films,chemicals removed from the films by the air stream weretrapped using a C18-bonded silica powder. Contamination ofthe C18 trap which may have occurred and was corrected forby taking C18 blanks. The C18 blank procedure involvedrinsing clean C18 with solvent then analyzing the solventusing the same procedure as a regular sample. Clean beadswere also used to calculate a film–air partition coefficient toanalyze for background values on the beads, and contam-ination that may occur during the generator columncollection (bead blank). The blank value for the air-side(Cair) was taken as the average of all solvent and C18 blanks.Film–air partition coefficients reported in this paper arecorrected for C18 and solvent blanks but not for bead blanks,under the assumption that all PCBs on the beads would beavailable for partitioning, regardless of their source.

3. Results and discussion

3.1. Quality control/quality assurance

To evaluate chemical recovery, the synthetic film ofEVA was spiked with a known solution of PCBs and OCPs.

y = 1.14x - 1.14r2 = 0.98

(Harner et al.)

y = 0.89x + 1.19r2 = 0.97

(Wilcockson and Gobas)

y = 0.98x + 0.55r2 = 0.84

(Wilford et al.)y = 0.82x + 1.81

r2 = 0.89(Current Study)

6

7

8

9

10

11

6 7 8 9 10 11log Koa

log

Kev

a-ai

r

Harner et al. Wilcockson et al. Wilford et al. Current study

Fig. 1. Results from various studies comparing the measured EVA–air partition coefficient to literature octanol–air partition coefficient for selected PCBs

and OCPs. Linear correlation was observed from all the independent studies.



The EVA was extracted and analyzed to calculate recoveryvalues. The overall average recovery for the 18 analyteswas 107712% (where n ¼ 4 at different concentrations ofthe spike solution ranging from 0.5 to 50 pg mL�1). Theaverage recovery for each compound ranged from 82.8% to119.8%. When rinsing the EVA coating from the glassbeads, two DCM washes were able to remove over 90% ofthe total EVA from the beads. Of the total EVA mass(combined from the two washes), the first and secondwashes removed approximately 70% and 30% of the totalfilm extracted, respectively. The residual mass wasdetermined by subsequent washes.

For the ambient surface film Kfa experiments, most ofthe C18 and solvent blanks had individual PCB levels belowthe lowest point of the calibration curve (0.1 pg mL�1).Congener concentrations were corrected by subtractingthe average blank values that were 40.1 pg mL�1. Themajority of the blank corrections were well below 25% ofthe amount in the air samples. The compounds with thelargest discrepancy were the higher molecular weightcongeners, mainly due to the very low levels detected.

The bead blank values were mainly below detectionalso, except for the case of the beads used in the PGB-3and Dorset studies, where blank values were higher thanin other studies, possibly due to residual compounds fromprevious usage of the beads. Blank values here (represent-ing �50 g of beads or �1500 individual beads) for thesame PCB congeners as analyzed on the ambient filmsamples, ranged from below the lowest calibration pointto over 700 pg per 50 g sample of beads (average of�320717 pg, n ¼ 3) in the summer, and from non-detected (ND) to over 900 pg per 50 g sample of beads(average of �3547159 pg, n ¼ 3) in the winter samplingperiod. As previously mentioned, for Kfa determinations,the bead concentrations were not blank-corrected underthe assumption that all PCBs in the presumed residual onthe beads and those deposited during deployment canpartition into the air, which may lead to a potential bias inthe partition values.

While field film samples may also contain interferencesthat have a similar retention time to PCBs on GC–ECDanalysis, further resolution by GC/MS was not required. Inlaboratory measurements of Kfa, only selected PCBs werepresent in the starting mixture, allowing easy resolutionand quantification of PCBs. While the field samples didexhibit higher baseline noise, PCB peaks were generally wellresolved and quantifiable, with the exception of the earliesteluting congeners. In the case of the KEVA–air values, previousexperiments have shown that EVA (and surface films) has arelatively higher capacity for non-polar hydrophobic com-pounds, such as PCBs, relative to the smaller sorptioncapacity for labile and polar compounds that may be foundin typical air samples (Harner et al., 2003; Farrar et al.,2005). Consequently, analysis by GC–ECD was adequate forPCB identification and quantification.

3.2. KEVA–air (EVA–air partition coefficient)

Fig. 1 summarizes the relationship between thelaboratory-measured EVA–air partition coefficient mea-

sured at 25 1C (KEVA–air) and the literature values of Koa of18 PCB congeners and organochlorine pesticides tempera-ture-corrected to 25 1C. The log–log relationship compareswell with previous measurements by Harner et al. (2003),Wilcockson and Gobas (2001) and Wilford et al. (2003).The slopes of near 1 from all four studies are an indicationthat EVA–air partitioning can be predicted based on theKoa for PCBs and for the organochlorine pesticides. A two-tailed statistical t-test performed on the slopes showedthat they were not statistically different at a 95%confidence level (p ¼ 1, 1, 0.53, 0.44, respectively, for thefour slopes). This is useful because Koa and its temperaturedependence have been measured for many chemicals (e.g.Harner and Mackay, 1995; Harner and Bidleman, 1996,1998). Results from the EVA work also serves as a test ofthe method used to examine ambient films.

Wilford et al. (2003) showed that similarly to Koa,KEVA–air also depends on temperature, where approxi-mately every 10 1C decrease in temperature results in anincrease of the partition coefficient by a factor of 3. Theenthalpies of phase change (DH) for KEVA–air and Koa aresummarized in Table 2, where DH was derived using theideal gas law. These results have implications for Kfa

values which, similar to Koa and KEVA–air, may also betemperature dependent.

3.3. Film– air partition quotients (Qfa) and partition

coefficients (Kfa)

3.3.1. Derived film–air partition quotients

Qfa values were calculated using the film PCB concen-trations from a field sample collected in December 2003where film was accumulated for �90 days (film PCBconcentration values were obtained as described in themethods section) and using gas-phase PCB air concentra-tion values collected over a 24 h period in downtownToronto at a site �200 m away at the Gage Building

Table 2Enthalpy of phase change (DH) for KEVA–air and Koa, where

log KEVA–air ¼ A+B/T, slope B ¼ DH/2.303R and the DH values are

literature-derived values (Harner and Bidleman, 1996; Harner and

Mackay, 1995)

Chemical A B DHEVA–air DHoa

PCB-3 �4.14 3395 65.0 66.4

a-HCH �1.67 2787 53.4 61.9

b-HCH 95

Lindane �0.67 3040 58.2 65.4

PCB-18 �1.83 2822 54.0 N/A

PCB-15 73

d-HCH 93

PCB-29 �1.42 2897 55.5 72.6

PCB-28 �1.36 2940 56.3

PCB-53 �1.55 2935 56.2 75.9

Heptachlor 0.86 2526 48.4 66.2

PCB-52 �1.92 3131 60.0 N/A

PCB-49 �2.02 3191 61.1 76.2

Aldrin �1.07 3097 59.3 71

PCB-61 �1.83 3365 64.4 66.3

PCB-66 73

Endo1 83

Endrin 85



(Environment Canada, 2003). Fig. 2 is a plot of log Qfa

versus log Koa for 15 PCB congeners, where Koa values inthis plot are corrected to 12.5 1C, the average temperatureover the sampling period using gas chromatographicrelative retention times (RRT), as described by Harjuet al. (1998).

If the film had the same partitioning properties asoctanol and film–air apparent equilibrium was establishedfor PCBs, the data would have a slope near unity and fallnear the 1:1 line. However, the regression line in Fig. 2 hasa slope of 0.46 with the lower molecular weight PCBs nearthe 1:1 line and the higher molecular weight PCBsdeviating below the line. Interestingly, Fig. 3 shows the

analogous log–log plot of Qfa and KEVA–air, suggests thatEVA may be a better surrogate for the surface film thanoctanol.

The better agreement with EVA may be the result of itscomposition being more similar to real surface films. Liuet al. (2003) found that the most abundant organiccompounds of those analyzed in surface films fromToronto were the carboxylic acids. Rogge et al. (1993)analyzed air particles in Los Angeles and had similarresults, although the proportions of monocarboxylic acidsand dicarboxylic acids were slightly different. Lam et al.(2005) found Toronto surface films to comprise �30%carbohydrates, 30% aliphatic and �20% aromatics. Esters(functional group of EVA) and carboxylic acids have closerpolarity values (based on the dielectric constant) thancarboxylic acids and alcohols (e.g. octanol), providing amechanistic explanation for the closer agreement of Qfa

with KEVA–air than Koa. It should be noted that the betteragreement between log Qfa with log KEVA–air (Fig. 3)compared to log Koa (Fig. 2) may also be due to thedifference in the data points, where the log Qfa withlog KEVA–air comparison is between OCPs and PCBs and thelog Qfa�log Koa comparison is between PCBs only.

Harner et al. (2003) previously examined uptake ofPCBs using EVA as a passive sampler of ambientcompounds in air. Results from their study showed thatuptake was air-side controlled and the time for filmconcentrations to respond to a new equilibrium (ex-pressed as t95 or time to 95% of the equilibrium value)depended on the film thickness and the capacity of thefilm for the chemical. Film capacity increases for highermolecular weight PCBs. For a relatively thin, 70 nm film(comparable to �30 days urban film growth, based on theaverage of the reported growth rates in Wu et al. (2008) of�2.2 nm day�1), t95 values range from 0.7 days for PCB-28to 22 days for PCB-153.

In the case of these derived quotients, factors that cancause differences in the slope are summarized by Simciket al. (1998), including different activity coefficients in theorganic matter content and changes in atmosphericconcentrations of compounds. For temperature-relatedimpacts, Pankow and Bidleman (1992) reported thatchanges in ambient temperature may lead to adsorptionor desorption of particle-bound analytes, resulting ingas–particle distributions that are not in equilibrium.Lohmann et al. (2000) also studied temperature effects onequilibrium partitioning. In the event of new sources ofrelatively clean particles (or film), the more volatile PCBswill re-equilibrate faster, changing the slope of the log Kp

(gas–particle partition coefficient) versus log Koa curve toa slope shallower than +1. Other explanations may be thathigher molecular weight PCB congeners have a greateraffinity for the film and will take longer to reach apparentequilibrium in the air or that octanol may not be anappropriate surrogate for ambient surface films. Analy-tical uncertainty may also be a factor.

3.3.2. Film–air partition coefficients (Kfa) from Downtown

Toronto (PGB-1 and PGB-2)

Several samples were chosen at random from PGB-1and PGB-2 studies to derive a film–air partition coefficient

y = 0.4614x + 4.1376r2 = 0.61

7

8

9

10

11

12

7 8 9 10 11 12log Koa

log

Qfa

1:1

Fig. 2. Film–air partition quotients (Qfa, calculated partition quotient

based on ambient gas-phase air concentrations of PCBs and film PCB

concentrations measured as part of this study) in relation to tempera-

ture-corrected literature octanol–air partition coefficient values.

y = 1.0609x - 1.1787r2 = 0.92

6

7

8

9

10

11

6 7 8 9 10 11log Keva-air

log

Qfa

Fig. 3. Film–air partition quotients (Qfa, calculated partition quotient

values based on ambient gas-phase air concentrations of PCBs and film

PCB concentrations measured as part of this study) derived for PGB-1

study and plotted against the log KEVA–air, all corrected to 22 1C.



(Kfa) in the laboratory chamber studies. Table 3 sum-marizes the sample parameters. The film evolution ratesfrom all the studies conducted at PGB were linear theentire length of the study, ranging from 1.8 to 3.1 nmday�1 (Wu et al., 2008). However, films collected atdifferent points in time are expected to have a variedchemical composition which may alter film–air partition-ing properties, as film–air equilibrium will also be in flux.

In the two samples chosen from PGB-1 (sample no. 8,film thickness �24 nm and sample no. 25, film thickness�179 nm), there was better correlation between the log Kfa

and log Koa values for the later sample with thicker filmthan the thinner film (r2

¼ 0.9422 for sample no. 25where slope is 0.96 versus r2

¼ 0.7777 in sample no. 8where slope is 0.75). Similar to previous results from theEVA study, thinner films may also respond more rapidly tochanges in ambient air concentrations, resulting in largerdeviations from apparent equilibrium for some congeners.As equilibrium for these surface films is a function of thefilm capacity (Koa), it will adjust to a new equilibriumquicker for the lower molecular weight compounds than itwill for the higher molecular weight compounds becauseof its greater capacity for the latter. This difference in timescales for equilibrium can lead to poor correlations whenmeasuring a broad range of PCB congeners.

In the PGB-2 study, three separate samples were usedto measure a film–air partition coefficient. A samplecollected near the midpoint of the study (sample no. 11, 84days of film accumulation, 242 nm film thickness) showedthe best correlation between the log Kfa and log Koa values.A sample from near the beginning of the study (sampleno. 5, 35 days of accumulation, 46 nm film thickness) hadthe shallowest slope; while a sample collected from nearthe end (sample no. 22, 210 days of film accumulation,307 nm film thickness) had a slope that fell somewherebetween the previous two. These results indicate thatthicker films do not always give better correlations oflog Kfa and log Koa. As discussed previously, variable filmcomposition with time is also important.

As shown in Fig. 3 for field-derived Qfa values (i.e.calculated with Cfilm and ambient Cair), in Fig. 4, the lowermolecular weight PCBs again fell closer to the 1:1 linethan the higher molecular weight congeners. Besides filmthickness and analytical errors associated with measuringthe mass and PCB content of thin films, the differentreasons for variations in slope as described previously forthe film–air partition quotients (Qfa) may be applied to the

current situation. However, as temperature is kept con-stant in controlled chamber experiments determining Kfa,this factor is unlikely to contribute to the uncertainty inthe equilibrium partitioning (Harner and Bidleman, 1996).

Changes in film composition amongst the samplingperiods constitute a possible explanation for the deviationof slopes. The variations in film composition result indifferent activity coefficients of compounds in the film, aparameter which is related to the solubility of eachcompound in the organic phase, or in this case, thesolubility of PCBs in the surface films. Previously pub-lished results (Tasdemir et al., 2004; Cotham and Bidle-man, 1995) stated that regression values between Kp

(gas–particle partitioning) and the log poL (subcooled

liquid vapor pressure) vary from one event to another.Chandramouli et al. (2003) have also reported that Kp issensitive to the aerosol mix from different sources, whichcan vary from one sample to another even over the courseof a single study period. Over the span of the fairly lengthyPGB-2 study (273 days), film composition would be moresubject to change as its components reacted and evolvedwith time. Meteorological factors and other inputs thatchange with season may also have contributed to thechanges observed in the film. The sum of these changes islikely to influence the overall partitioning of PCBs.

Goss and Schwarzenbach (2001) reinforced the pointthat a slope deviating from the 1:1 line is not necessarilyan indication of non-equilibrium between phases. Linearfree energy relationships (e.g. double logarithmic correla-tions, as explored in this paper) are only capable ofdescribing compound variability within a single substanceclass and do not account for differences in the organicphase. While log–log relationships of Kfa and Koa do notaccount for numerous factors, this approach is useful inthe development of empirical relationships, especially inthe application of these relationships to models thatinvestigate the contribution of the surface organic film tocontaminant cycling in urban environments.

3.4. Film– air partition coefficients (Kfa) from the

urban–rural study (PGB-3 and Dorset)

Beads collected at the paired urban–rural sites (PGB-3and Dorset) were analyzed for film mass and PCB content.The results are presented in Wu et al. (2008). Filmaccumulation was found to be higher at the urban sitethan at the rural site for both winter and summer studies.Beads were collected once at the end of each study(Table 1) and were used to determine experimental Kfa

values in order to verify whether urban and rural surfacefilms exhibited different film–air partitioning propertiesfor PCBs.

The PCB film concentrations calculated for the ruralstudy were based on the average film mass and averagePCB mass obtained from two bead samples from the site.These duplicates tended to be highly variable, especiallyfor the rural site, which had little film mass. The resultingslopes (where Kfa was calculated using the larger mass ofthe two samples) were shallow, from �0.03 to 0.28 for therural film and from 0.16 to 0.34 for the urban film (Fig. 5).

Table 3Summary of the samples chosen from PGB-1 and PGB-2 for Kfa

laboratory experiments

Study Sample

no.

Mean ambient

temperature (1C)

Days of film

accumulation

Film

thickness

(nm)

Slope of Kfa

versus Koa

PGB-

1

8 16.8 20 23.7 0.75

25 3.5 71 178.6 0.96

PGB-

2

5 15.5 35 45.9 0.10

11 15.3 84 242.6 0.72

22 �3.4 210 306.8 0.47



The shallower slopes from the PGB-3 beads compared toPGB-1 or PGB-2 study are difficult to explain although ashallow slope of �0.1 was also observed for one of thePGB-2 samples. These results seem to indicate that Koa

may not be the most suitable parameter for approximat-ing film–air partitioning because ambient surface filmsmay vary, leading to different partitioning behavior,depending on the film composition, or more informationis required on the organic fraction of aerosols during thestudy period.

The film thicknesses recorded at the urban siteaveraged �158 and �241 nm, respectively, in winter andspring/summer. At the rural site, the films were thinnerwith an average of �5 nm in the winter and �20 nm in thespring/summer. While the surface film can be measuredin rural sites, its relative thinness in comparison to theurban films makes its impact on rural cycling of

contaminants less important than in an urban setting.Also, impervious surfaces are more abundant in urbancenters, increasing the potential for film accumulation.Other organic phases, such as vegetation, would undoubt-edly play a greater role in contaminant dynamics than thesurface film.

Blanchard et al. (2002) analyzed the chemical compo-sition of the organic fraction of atmospheric aerosols at anurban site and compared it to a rural site. The concentra-tions of various chemicals were similar between the twosites. However, the chemical profiles for alkanoic acidsand n-alkanes were different, with a greater emphasis onbiogenic sources in the rural site, and anthropogenicsources in the urban site. The organic carbon to elementalcarbon ratio also differed between the two sites, againwith the rural site showing a greater potential foraccumulation of particles of biogenic origin. The differ-ences in the particle composition and source between

y = 0.9572x - 0.8859r2 = 0.94

y = 0.7511x + 1.7476r2 = 0.78

6

7

8

9

10

11

Literature Log Koa Values

Log

Kfa

Sample #8 (~24 nm) Sample #25 (~179 nm)

y = 0.4655x + 4.3391r2 = 0.73

y = 0.1035x + 8.1996r2 = 0.13

y = 0.7203x + 0.8891r2 = 0.77

6

7

8

9

10

11

6 7 8 9 10 11

6 7 8 9 10 11

Literature Log Koa Values

Log

Kfa

Sample #5 (~46 nm) Sample #11 (~243 nm)

Sample #22 (~307 nm)

1:1

1:1

Fig. 4. Laboratory-measured film–air partition coefficients (Kfa) from

PGB-1 (a) and PGB-2 (b) studies, plotted against log Koa for selected PCBs

(congeners 8, 18, 28, 52, 49, 101, 118, 153, 105, 137, 138, 156, 171, 180, 209).

y = 0.339x + 4.4495r2 = 0.71

y = -0.0325x + 8.9907r2 = 0.037

6

7

8

9

10

11

6 7 8 9 10 11Literature Log Koa

Log

Kfa

PGB Winter Dorset Winter

y = 0.1607x + 5.7315r2 = 0.37

y = 0.2847x + 5.2126r 2 = 0.79

6

7

8

9

10

11

6 7 8 9 10 11Literature Log Koa

Log

Kfa

PGB Spring Dorset Spring

1:1

1:1

Fig. 5. Laboratory-measured film–air partition coefficients (Kfa) from

PGB-3 and Dorset winter (a) and spring/summer (b) studies for selected

PCBs (congeners 8, 18, 28, 52, 49, 101, 118, 153, 105, 137, 138, 156, 171, 180,

209). The rural films have greater partitioning variability, which may be

partly due to lower levels of film measured.



an urban and a rural site may also be reflected in thepartitioning behavior of chemicals with the surface filmsand ambient air between the two sites.

Because of the multiple factors that may be influentialin determining film–air partitioning, future work canfocus on isolating some of the most relevant. This mayinvolve a more in-depth investigation into the physicalattributes of the film, for example, film orientation andstructure, as this may affect air–surface partitioning andwash-off.

4. Implications

Results from this study showed that both field-derivedpartitioning quotients (Qfa, calculated values based onfield measurements) and experimentally derived partitioncoefficients (Kfa, derived from chamber studies) showed aclear but variable correlation in log–log plots against Koa;regression slopes were variable and generally closer tounity (1:1) when r2 values were higher. For the fieldvalues, departure of Qfa away from the 1:1 line wasobserved and reasons for this deviation may be theaffinity of the film for high molecular weight PCBs. Kfa

values derived from controlled chamber experimentsshowed considerable variability especially in the slope oflog Kfa versus log Koa deviating from unity. By analogy toprevious research on gas–particle partitioning, thesedeviations may be reflecting changes in the surface filmcomposition, which alter the activity coefficient of PCBs inthe organic fraction. The correlation between Kfa ofcollected films to both the octanol–air partition coefficient(Koa) as well as the gas–particle partitioning value (Kp) is auseful relationship in the development of empiricalapproaches to quantify the contribution of the surfaceorganic film to ambient air concentrations. While therelationship between Koa and Kfa is variable, it is possibleto obtain average relationships between the two para-meters that would allow for the inclusion of surface filmsin urban fate models. Some of the observed variation maybe due to changes in the film composition from one day tothe next, further research needs to focus on variations infilm composition over a period of time. Additional work isalso required to look at film–air partitioning dependencyon temperature and other influencing factors, such as thecontinual washoff due to precipitation and subsequentregeneration.

The results from the EVA studies may also suggest thatEVA is a more suitable surrogate for ambient surface films.However, as the Kfa values in this study are calculatedbased on bulk film, and not the organic fraction (which isthe fraction into which organic contaminants wouldpartition), it is important to evaluate partitioning on thisbasis also. This is an area for future work.

Acknowledgments

The authors would like to acknowledge Jennifer Truongfor invaluable assistance in sample collection, and CraigButt for help in experimental setup and method develop-ment. Much of the experimental EVA work was performed

by Andy Skinner. This study was funded by an NSERCStrategic Grant (Project no. STP-258056). Gas-phase PCBdata used in the derivation of film–air quotients wasobtained from Environment Canada’s Integrated Atmo-spheric Deposition Network.

References

Blanchard, P., Brook, J.R., Brazal, P., 2002. Chemical characterization ofthe organic fraction of atmospheric aerosol at two sites in Ontario,Canada. Journal of Geophysical Research 107, 8348.

Boyd, M.J., Bufill, M.C., Knee, R.M., 1993. Pervious and impervious runoffin urban catchments. Hydrological Sciences Journal 38, 463–478.

Butt, C.M., Diamond, M.L., Truong, J., Ikonomou, M.G., Ter Schure, A.F.H.,2004. Spatial distribution of polybrominated diphenyl ethers inSouthern Ontario as measured in indoor and outdoor windoworganic films. Environmental Science and Technology 38, 724–731.

Chandramouli, B., Jan, M., Kamens, R.M., 2003. Gas–particle partitioningof semivolatile organic compounds (SOCs) on mixtures of aerosols ina smog chamber. Environmental Science and Technology 37,4113–4121.

City of Toronto, 2005. City of Toronto: Toronto facts. [Online] URL:/www.toronto.ca/toronto_facts/geography.htmS (accessed 29.10.05).

Cotham, W.E., Bidleman, T.F., 1995. Polycyclic aromatic hydrocarbons andpolychlorinated biphenyls in air at an urban and rural site near lakeMichigan. Environmental Science and Technology 29, 2782–2789.

Diamond, M.L., Gingrich, .E., Fertuck, K., McCarry, B.E., Stern, G.A., Billeck,B., Grift, B., Brooker, D., Yager, T.D., 2000. Evidence for organic film onan impervious urban surface: characterization and potential terato-genic effects. Environmental Science and Technology 34, 2900–2908.

Diamond, M.L., Priemer, D.A., Law, N.L., 2001. Developing a multimediamodel of chemical dynamics in an urban area. Chemosphere 44,1655–1667.

Environment Canada Integrated Atmospheric Deposition Network, 2003.[Online] URL: /http://www.msc-smc.ec.gc.ca/iadn/index_e.htmlS(accessed 26.09.04).

Farrar, N.J., Harner, T., Shoeib, M., Sweetman, A., Jones, K.C., 2005. Fielddeployment of think film passive air samplers for persistent organicpollutants: a study in the urban atmospheric boundary layer.Environmental Science and Technology 39, 42–48.

Finizio, A., Mackay, D., Bidleman, T., Harner, T., 1997. Octanol–airpartition coefficient as a predictor of partitioning of semi-volatileorganic chemicals to aerosols. Atmospheric Environment 31,2289–2296.

Gingrich, S.E., Diamond, M.L., Stern, G.A., McCarry, B.E., 2001. atmo-spherically derived organic surface films along an urban–ruralgradient. Environmental Science and Technology 35, 4031–4037.

Goss, K.-U., Schwarzenbach, R.P., 2001. Linear free energy relationshipsused to evaluate equilibrium partitioning of organic compounds.Environmental Science and Technology 35, 1–9.

Harju, M.T., Haglund, P., Naikwadi, K.P., 1998. Organohalogen Compounds35, 111–114.

Harner, T., Bidleman, T.F., 1996. Measurements of octanol–air partitioncoefficients for polychlorinated biphyenyls. Journal of Chemical andEngineering Data 41, 895–899.

Harner, T., Bidleman, T.F., 1998. Measurements of octanol–air partitioncoefficients for polycyclic aromatic hydrocarbons and polychloro-napthalenes. Journal of Chemical and Engineering Data 43, 40–46.

Harner, T., Mackay, D., 1995. Measurement of octanol–air partitioncoefficients for chlorobenzenes, PCBs and DDT. EnvironmentalScience and Technology 29, 1599–1606.

Harner, T., Farrar, N.J., Shoeib, M., Jones, K.C., Gobas, F.A.P.C., 2003.Characterization of polymer-coated glass as a passive air sampler forpersistent organic pollutants. Environmental Science and Technology37, 2486–2493.

Lam, B., Diamond, M.L., Simpson, A.J., Makar, P.A., Truong, J., Hernandez-Martinez, N.A., 2005. Chemical composition of surface films on glasswindows and implications for atmospheric chemistry. AtmosphericEnvironment 39, 6578–6586.

Law, N.L., Diamond, M.L., 1998. The role of organic films and the effect onhydrophobic organic compounds in urban areas: an hypothesis.Chemosphere 36, 2607–2620.

Liu, Q.-T., Chen, R., McCarry, B.E., Diamond, M.L., Bahavar, B., 2003.Characterization of polar organic compounds in the organic film onindoor and outdoor glass windows. Environmental Science andTechnology 37, 2340–2349.



Lohmann, R., Harner, T., Thomas, G.O., Jones, K.C., 2000. A Comparativestudy of the gas–particle partitioning of PCDD/Fs, PCBs and PAHs.Environmental Science and Technology 34, 4943–4951.

Pankow, J.F., Bidleman, T.F., 1992. Interdependence of the slopes andintercepts from log–log correlations of measured gas–particlepartitioning and vapor pressure—I. Theory and analysis of availabledata. Atmospheric Environment 26A, 1071–1080.

Priemer, D.A., Diamond, M.L., 2002. Application of the multimediaurban model to compare the fate of SOCs in an urban andforested watershed. Environmental Science and Technology 36,1004–1013.

Rogge, W.F., Mazurek, M.A., Hildemann, L.M., Cass, G.R., Simoneit, B.R.T.,1993. Quantification of urban organic aerosols at a molecular level:identification, abundance and seasonal variation. AtmosphericEnvironment 27A, 1309–1330.

Simcik, M.F., Franz, T.P., Zhang, X., Eisenreich, S.J., 1998. Gas–particlepartitioning of PCBs and PAHs in the Chicago urban and adjacent

coastal atmosphere: states of equilibrium. Environmental Scienceand Technology 32, 251–257.

Tasdemir, Y., Vardar, N., Odabasi, M., Holsen, T.M., 2004. Concentrationsand gas–particle partitioning of PCBs in Chicago. EnvironmentalPollution 131, 35–44.

Wilcockson, J.B., Gobas, F.A.P.C., 2001. Thin-film solid-phase extraction tomeasure fugacities of organic chemicals with low volatility inbiological samples. Environmental Science and Technology 35,1425–1431.

Wilford, B., Skinner, A., Harner, T., Jones, K.C., 2003. Measurements oftemperature dependent film/air partition coefficients of POPs. In:Proceedings of the 24th Annual Meeting of the Society of Environ-mental Toxicology and Chemistry (SETAC) in North America, Austin,TX, USA. Poster presentation.

Wu, R.W., Harner, T., Diamond, M.L., 2008. Evolution rates and PCB contentof organic films that develop on impervious urban surfaces. Atmo-spheric Environment, in press, doi:10.1016/j.atmosenv.2008.01.066.


Partitioning characteristics of PCBs in urban surface films

Documents