Science of the Total Environment 472 (2014) 13–19
Contents lists available at ScienceDirect
Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv
Ambient polychlorinated biphenyl levels and their evaluation in a metropolitan city S. Levent Kuzu ⁎,1, Arslan Saral 1, Gülsüm Summak 1, Hatice Çoltu 1, Selami Demir 1 Yildiz Technical University, Civil Engineering Faculty, Environmental Engineering Department, 34220, Davutpaşa-Esenler, Istanbul, Turkey
H I G H L I G H T S • PCB concentrations are found to be correlated to mixing height during summer period. • MLR, PSCF and gas/particle particle partitioning results suggested the same outcome as evaporation of PCBs from sea. • PSCF input parameters were normalized according to mixing height for the first time.
a r t i c l e
i n f o
Article history: Received 26 June 2013 Received in revised form 4 November 2013 Accepted 5 November 2013 Available online 27 November 2013 Keywords: PCB concentrations MLR PSCF Gas/particle partitioning
a b s t r a c t In this study, summer and autumn ambient PCB concentrations were investigated in metropolitan city of Istanbul. 84 congeners were targeted from di-CBs to nona-CBs on both particle and gaseous phases. Gaseous ambient concentrations were determined to be 372 ± 134 pg·m−3, while on the particle phase this value was 49 ± 17 pg·m−3, corresponding to an average of 420 pg·m−3. About one-tenth of all PCBs lay in ambient aerosols, while 90% of all comprise 2-, 3-, 4-, and 5-CBs. Measured ambient concentrations of each congener group were tested against meteorological data. The di-CB concentrations were independent of ambient temperature while northerly winds lead to an increase in their concentrations, which was an indicator of considerable contribution to di-CB concentrations from the medical waste incineration plant in Istanbul. In contrast, other congeners' concentrations were found to be correlated with southerly winds. Being an inland sea and having been contaminated, for years, by industrial discharges along the coastline, volatilization from Marmara Sea was considered as the most probable source of other congeners. PSCF analysis was run with 12-hour trajectories to locate possible local sources and check these results. Gas/particle partitioning was applied using three different models. mr and br values for log P0L model were determined as −0.23 ± 0.09 and −3.25 ± 0.38, respectively. For absorption based log Koa model, m and b values were calculated as 0.23 ± 0.08 and −4.73 ± 0.83, respectively. © 2013 Elsevier B.V. All rights reserved.
1. Introduction Of the chlorinated benzene rings, PCBs pose a great threat to human health as well as animals, plants and so forth. For this reason, these semi-volatile organic compounds (SVOCs) have attracted much interest especially in the last two decades (Cindoruk and Tasdemir, 2010). Earlier, PCBs were used as dielectric fluids, lubricants, hydraulic and heat transfer fluids until their ban in 1970s. Currently, major sources of atmospheric PCBs are listed as leakage from previously contaminated fluids, open burning, waste incineration, evaporation from contaminated sites, landfills and sludge drying beds (Breivik et al., 2002; Hsu et al.,
⁎ Corresponding author. E-mail addresses: [email protected] (S.L. Kuzu), [email protected] (A. Saral), [email protected] (G. Summak), [email protected] (H. Çoltu), [email protected] (S. Demir). 1 Tel: +90 212 383 53 78; fax: +90 212 383 53 58. 0048-9697/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.scitotenv.2013.11.031
2003). Apart from these closed or semi closed systems, some production activities and combustion processes produce PCB emissions as byproduct. In an inventory study, considering each production type which has PCB emission factor, 920 kg·yr− 1 of industrial PCB release to atmosphere is estimated within Turkey (Kuzu et al., 2013). Among the production activities, steel production by electric arc furnaces with pre-heating has the highest emission factor (Odabasi et al., 2009). Once these compounds are emitted from the source, they tend to partition at different media due to their semi-volatile characteristics (Atkinson, 1991). However, equilibrium is never reached and a continuous transfer of PCBs takes place between distinct media due to dynamic structure of the atmosphere and other compartments of the Earth. Thus, researchers inclined to understand the behavior of the compounds in the environment (Carlson and Hites, 2004). Transport of these compounds in the atmosphere is the major pathway from the point of emission until their deposition (Chen et al., 2009).
14
S.L. Kuzu et al. / Science of the Total Environment 472 (2014) 13–19
Further, these compounds reach water and soil bodies through atmospheric deposition mechanisms (Birgul and Tasdemir, 2011). Istanbul is located on both European and Asian sides, separated by the Bosphorus Strait. The city is surrounded by the Black Sea on the northern side and Marmara Sea on the southern side. Istanbul is on the route of dominant atmospheric motion through which contaminants are transported over Europe, Asia and Mediterranean region (Karaca et al., 2009). The main aims of this study is to i) investigate the ambient PCB levels in this thirteen-million-city with no previous reports on ambient levels of PCBs, ii) identify potential source regions to the ambient PCB levels, iii) determine gas/particle partitioning by present models and discuss their results. 2. Materials and methodology 2.1. Sampling Sampling was conducted within Davutpasa Campus of Yildiz Technical University in Istanbul, Turkey (41°01′26″ N and 28°53′16″ E). The campus area is surrounded by the main coach station of Istanbul on the northern side, O-1 highway on the southern side where traffic density is more than 100,000 cars per day (Onat and Stakeeva, 2013), and a great number of light and heavy industrial facilities from various sectors on western and eastern sides. A high volume air sampler (HVAS) was used to collect both particle and gaseous phase pollutants (Tecora ECHO HiVol). HVAS was equipped with a filter holder and cartridge. Ambient air was first drawn through glass fiber filter (GFF) of 10 cm in diameter and then through poly urethane foam (PUF) cartridge consisting of two identical foam layers of 5 cm in diameter and 6 cm in thickness. The foam layers were in series. The sampler was operated at 350 L·min−1 for 24 h which corresponds to approximately 500 m3 of ambient air sample. The device was calibrated before each sampling campaign. Samples were collected from May 2012 to October 2012, representing summer and autumn seasons. Davis Vantage 2 Pro online weather station was employed to obtain meteorological data during the sampling. The mixing height data were gathered from Atmospheric Research Laboratory (NOAA, 2012). Gravimetric measurements were conducted by AND GR-202 analytical microbalance. Before sampling process, GFFs were wrapped with aluminum foil and heated at 450 °C in a furnace for 6 h to remove any organic residuals. Then, the filters were conditioned in a desiccator at a temperature of 20 ± 2 °C and a relative humidity of 50 ± 5%.
Samples were then cleaned up and fractionated through a column of 3 g silicic acid (3% water), 2 g neutral alumina (6% water), and 2 g Na2SO4 (Falconer et al., 1995; Tasdemir et al., 2004), which was prewashed with 20 mL of DCM and 20 mL of PE, respectively. Next, the sample in 2 mL of hexane was poured to the column and PCBs were eluted with 20 mL of PE. The procedure to switch the solvent was repeated and the final extract was taken into hexane again. These extracts were cleaned with H2SO4 and concentrated to 1 mL by stripping with ultra pure nitrogen gas. At the last stage internal standard mixtures of PCB#30 and PCB#121 of 5 ng·ml−1 were included for volume correction purpose. Quantification of PCB concentrations was performed by a GC-ECD system (Perkin Elmer Clarus 500) equipped with a HP-5MS (30 m × 0.25 mm × 0.25 μm), capillary column. The column temperature was kept constant at 70 °C for 2 min and then raised to 150 °C at a rate of 25 °C·min−1, to 200 °C at a rate of 3 °C·min−1, and finally to 280 °C at a rate of 8 °C·min−1. The temperature was kept constant at 280 °C for 10 min. The inlet temperature 250 °C and detector temperature was 320 °C. Ultra pure helium was used as the carrier at a flowrate of 1.2 mL·min−1. The makeup gas (nitrogen) flowrate was 25 mL·min −1 . No split flow was used during the analysis. The calibration was accomplished using seven standard solutions with concentrations ranging from 0.1 ng·μL−1 to 30 ng·μL−1 After each 10 sample injections, stability was checked with the medium standard. The average value of coefficients of determination for distinct PCB congener was approximately 0.995. The investigated 84 congeners are those presented in Fig. 4 and additionally, PCB#7/9, PCB#6, PCB#12/13, PCB#18, PCB#100, PCB#66/95, PCB#92, PCB#123. 2.3. Quality assurance/quality control (QA/QC) All samples were spiked with surrogate standards and the recovery efficiencies are shown in Table 1. Least volatile congeners have higher recovery efficiencies for both PUF plugs and TSP filters. Usually PUF plugs had higher recovery efficiency among the same congener group. Limit of detection (LOD) for each PCB congener was assumed as its blank signal plus three times its standard deviation (Cindoruk and Tasdemir, 2007; Odabasi et al., 1998). Obtained LOD values for individual PCB congeners varied between 0 and 1.6 pg. Sample concentrations under LOD value were ignored. Blank samples were collected for each set of analysis. All results were blank corrected. 3. Results and discussion
2.2. Extraction and analysis
3.1. Ambient PCB concentrations
Following the sample collection, the samples were taken to the laboratory for gravimetric analysis and extraction. PCB surrogate standard mixture of PCB #14, #65, and #166 (5 ng·mL−1 for each congener) was added to the PUF cartridges and GFFs prior to the extraction procedure. The PUF cartridges were extracted with 1:1 (v/v) acetone (ACE)/ hexane (HX) mixture for 24 h in a Soxhlet extractor (Cetin et al., 2007). GFF contents were extracted in an ultrasonic bath with a mixture of dichloromethane (DCM)/petroleum ether (PE) of 1:4 (v/v) (Cindoruk and Tasdemir, 2010). Following the extraction procedure, filters were soaked in a glass jar capped with Teflon lined lids containing 25 mL of solvent mixture for 30 min. A second aliquot was formed with the same amount of solvent mixture to continue extraction for another 30 min. At the end of 1 h, the solvent mixture was evaporated and the volume was reduced to 5 mL in a rotary evaporator, which is followed by the switching of the solvent mixture to hexane by the addition of 15 mL of hexane. The latter was repeated twice. The final mixture was stripped gently by constant flow of ultra-pure nitrogen gas until its volume reached to 2 mL.
Both gaseous- and particulate-phase PCB concentrations in ambient air were measured in Istanbul between May 2012 and November 2012. Results are shown in Fig. 1 along with the atmospheric mixing heights during the sampling periods. Average gaseous phase PCB concentrations were found as 372 ± 134 pg·m−3 whereas particle phase PCB concentrations were 49 ± 17 pg·m−3. Total ambient PCB concentration was 420 pg·m−3 on average. About 12% of the total PCB
Table 1 Recoveries of the surrogate standards. Statistical parameters Recoveries of TSP filters
Recoveries of PUF plugs
PCB#14 PCB#65 PCB#166 PCB#14 PCB#65 PCB#166 Average Standard deviation 75th percentile Median 25th percentile
78.8 13.3 85.6 78.3 70.2
81.2 12.3 87.8 85 75.5
88.6 14.4 95.2 89.4 81.7
79.5 15.4 86.9 77.6 69.3
80.2 12.8 86.5 83.4 73.1
96.6 14.5 106.5 93.5 85.8
S.L. Kuzu et al. / Science of the Total Environment 472 (2014) 13–19
15
Gas
Particle
30
Percentage (%)
25 20 15 10 5 0 2-CB
3-CB
4-CB
5-CB
6-CB
7-CB
8-CB
9-CB
PCB homologues Fig. 2. Homologue PCB distribution on particle and gaseous phase.
et al., 1997; Sofuoglu et al., 2004) so far to relate the meteorological data with SVOC concentrations. Fig. 1. Ambient PCB concentrations.
concentrations lay in particulate phase. Total ambient concentrations and their distribution in particulate and gaseous phases showed good agreements with the results of a previous study in Bursa, Turkey which is located about 80 km south of Istanbul (Cindoruk and Tasdemir, 2010). Further, similar results were obtained in South Korea (Yeo et al., 2003). There hadn't been any published papers reporting PCB concentrations in Istanbul before. PCB concentration values were available for four different cities in Turkey. An average of 105 pg·m−3 PCB concentration for 14 congeners was reported in Bolu located to the Northwestern part of Turkey (Yenisoy-Karakas et al., 2012). Cetin et al. (2007) reported results between 314 and 3136 pg·m−3 for 41 congeners at an urban and industrial site, respectively, in Aliaga, İzmir. In Bursa (Cindoruk and Tasdemir, 2010), the concentrations were ranged between 316 and 570 pg·m−3 for 83 congeners. This study was conducted at four different sites representing coastal, urban, semi-rural and residential sites. Results were not differed as much as in Aliaga. The last available concentrations were measured in Konya, located in the middle of Turkey. Ozcan and Aydin (2009) observed an average concentration of 106 pg·m−3 at urban site. Regression was performed to ambient gaseous phase concentrations and mixing height values to checked whether the concentrations are correlated with mixing heights or not. However, it was discovered that results are not correlated (p b 0.15). As a next step data sets were divided into two groups as summer and fall data. Regression was again performed for each data group. Summer results were borderline statistically significant (p b 0.07), whereas fall results were statistically insignificant (p b 0.46). Atmospheric concentrations are likely to be driven by a continuous source during the summer period. It could be deduced that evaporation from previously contaminated sites such as soil or water surface could be the major PCB source of gaseous phase concentrations in warmer seasons. PCB homologue groups were illustrated for gaseous and particle phases in Fig. 2. Di-CBs to penta-CBs constitute approximately 87% of the total distribution. The distribution of di-CBs to penta-CBs for gaseous phase and particle phase was discovered to be 89% and 73%, respectively. Higher chlorinated biphenyls tend to be present in the particle phase.
3.2. Relationship between meteorological parameters and gaseous-phase concentrations In order to investigate the effect of meteorological conditions on PCB concentrations, multiple linear regression (MLR) was performed. MLR has been utilized by several researchers (Bozlaker et al., 2008; Hillery
ln P ¼ m1
. 1
T
þ m2 ðuÞ þ m3 ð cosWDÞ þ constant:
ð1Þ
In Eq. (1), the terms P, T, u and WD are corresponding to gas-phase partial pressure (atm), average ambient temperature (°K), wind speed (m·s− 1) and prevailing wind direction (degrees), respectively. m1, m2, and m3 are the regression parameters. Determined regression parameters are listed in Table 2. m1 values are proportional with the reciprocal temperature which means the higher temperature, the higher PCB concentration. m2 indicates that PCB concentrations increase when wind is blowing at slower speed. Negative m3 values are corresponding to southerly directions while positive values are corresponding northerly directions. Except di-CBs, vast majority of PCB homologue group concentrations are increased with the increasing temperature. The dependence of concentration to temperature may be attributed to the evaporation of previously deposited PCBs from terrestrial regions. Negative m2 values show that at higher wind speeds PCBs are swept by advection, consequently lower concentration is observed. The insignificance of di and tri CBs can be explained by local sources. A medical waste incineration plant is located at 20 km north of the sampling location. Jansson et al. (2011) expressed that flue gas from incineration plants are dominated by the least chlorinated PCB congeners. Moreover according to m3 value higher concentrations are observed when wind blows from the North. Finally, according to statistical results, it can be concluded that di-CBs are not related with meteorological parameters. According to m3 value, highest contribution in terms of wind direction occurs from southerly direction. Marmara sea is to the south of the sampling location where it is thought to be a polluted inland sea from the industrialized cities on the vicinity along the Marmara coastline (Gunindi and Tasdemir, 2010). Some other researchers stated that marine traffic is also accompanying to industrial SVOCs which is being transported from surrounding cities (Coelhan et al., 2006; Okay et al., 2009). Bursa is located to the south of the Marmara Sea. In this city,
Table 2 Multiple linear regression parameters.
Di-CBs Tri–CBs Tetra-CBs Penta-CBs Hexa-CBs Hepta-CBs Okta-CBs Nona-CBs ΣCBs
m1
m2
m3
r2
p1
p2
p3
303 −2259 −8528 −6505 −9522 −8628 −11,202 −7143 −4465
−0.06 −0.34 −0.26 −0.29 −0.35 −0.35 −0.32 −0.10 −0.25
0.06 −0.43 −0.90 −0.28 −0.19 −0.24 0.11 −0.35 −0.48
0.01 0.38 0.55 0.41 0.30 0.30 0.19 0.23 0.42
0.95 0.54 0.04 0.03 0.04 0.05 0.08 0.12 0.16
0.78 0.03 0.11 0.02 0.05 0.05 0.20 0.55 0.06
0.84 0.06 b0.01 0.11 0.45 0.34 0.76 0.19 0.02
16
S.L. Kuzu et al. / Science of the Total Environment 472 (2014) 13–19
Cindoruk and Tasdemir (2010) stated that PCBs are likely to be transported from Marmara Sea according to MLR results. They concluded that evaporation from sea surface and air–water exchange could be an active mechanism. Considering our results, this fact can be confirmed and it is observed that Marmara sea is acting as a PCB source. 3.3. PSCF analysis PSCF analysis was executed for a further evaluation of the previous outcomes of MLR. PSCF can be briefly described as conditional possibility, characterizing the spatial distribution of possible geophysical source locations made out by using trajectories reaching to the sampling site (Wimolwattanapun et al.). Via this method, probability field is developed which equates excessive concentration values at the receptor point with their probable source locations (Uygur et al., 2010). In the theory of the model, it is supposed that the end point of a trajectory in a grid cell is assumed to gather the pollutant in concern. After that, the incorporated pollutant is supposed to be transported to the sampling point along the trajectory pathway with the air parcel (Heo et al., 2009). PSCF can be calculated using Eq. (2): PSC F i; j ¼
M i; j Ni; j
ð2Þ
where ΣNi,j is the total number of trajectory segment endpoints in the ijth grid cell during the whole sampling period and ΣMi,j is the total number of trajectory segment endpoints in the same grid cell which is corresponding to samples that are exceeding the threshold criteria. Small ni,j values cause elevated PSCF results with high uncertainties. The small values usually encountered at more distant grids from the source due to sparse distribution of the trajectories. To reduce the effect of small ni,j values PSCF result multiplied by an arbitrary weighting function Wi,j (Polissar et al., 1999). Assigned Wi,j values are listed in Eq. (3).
W i; j ¼
8 > > < > > :
1:00 80bni; j 0:70 20bni; j ≤80 0:42 10bni; j ≤20 0:05 ni; j ≤10
:
ð3Þ
It was told previously that, in the present study PSCF analysis was executed to support the MLR results. According to MLR results two sources are likely to contribute ambient PCB concentrations i) Marmara Sea to the south and ii) medical waste incineration plant to the north. For that reason back-trajectories were calculated for 12 h. This selection
leads us to resolve short-range sources. The grid cells are formed with a resolution of 0.5° × 0.5° between longitudes of 25 E and 35 E and latitudes of 40 N and 44 N. A threshold criteria need to be chosen to attain some of the trajectories as polluted. Conventionally, average value of pollutant concentrations is regarded as the threshold. However, we think that, this doesn't reflect the real case due to dynamic structure of the atmosphere. A situation of high concentration and low mixing height doesn't mean that the corresponding trajectory at the time of sampling period induced the excessive pollutant amount. In contrast, any case of concentration value that is slightly less than threshold value with high mixing height doesn't mean that this trajectory should not be selected as polluted. Therefore, we decided to choose the threshold criteria as normalized ambient PCB concentrations according to atmospheric planetary boundary layer values. For the calculation, PCB concentrations are multiplied by mixing heights reported for the corresponding sampling dates. PCB loads per unit area were determined as a result of this normalization process. Trajectories were selected considering the mean value and the criterion value was set to the 75th percentile. Results are depicted in Fig. 3. TrajStat v.1.2.2.6 was used for calculation and mapping purpose (Wang et al., 2009). Considering the factors, two prevailing regions are likely to affect PCB concentrations in Istanbul. One of them is the Marmara Sea as stated before while the other one seems as Black Sea at the North. During the warmer seasons volatilization from water surfaces can be regarded as an important PCB source (Agrell et al., 1999). Even, there is a concern on the occurrence of PCBs at the marine environment (Li et al., 2012). Also, it should be pointed out that northerly trajectories reach to the sampling point passing over the medical waste incineration plant. As a result, MLR results seem to be in accordance with the PSCF results. 3.4. Gas/particle partitioning Particle and gas partitioning are crucial in determining the fate of chemicals in the atmosphere (Hoff et al., 1996; Yeo et al., 2003). Several models have been developed so far to predict the gas/particle partitioning. One of these models is the Junge–Pankow adsorption model. This model was developed according to linear Langmuir isotherm (Tasdemir et al., 2004). In the model it is supposed that SVOC compounds are adsorbed to active sites on the surface of the particle (Harner and Bidleman, 1998). Junge–Pankow model is shown in Eq. (4). ∅¼
cθ p0L þ cθ
Fig. 3. PSCF results during the sampling period.
ð4Þ
S.L. Kuzu et al. / Science of the Total Environment 472 (2014) 13–19
∅¼
ð5Þ
Ratio of measured and predicted values according to Junge–Pankow model were plotted in Fig. 4. For low molecular weighted (MW) congeners, measured values are higher than the predicted values. The model underestimated MW congeners. However an opposite case is present for higher MW congeners. In that case, model results over estimate the partitioning fraction. Observed results are consistent with the previous studies in Chicago and Bursa atmosphere (Tasdemir et al., 2004; Cindoruk and Tasdemir, 2007). Lee and Jones (1999) stated that, c and θ values, which are assumed, are effective in achieving over and under estimation. In addition, black carbon particles affect gas/particle partitioning (Lohmann and Lammel, 2004). These contributions are thought to be effective in deviating the values from 1 in Junge–Pankow adsorption model. Another model was applied in which gas/particle partitioning coefficient (Kp, m−3·μg) is plotted against supercooled vapor pressure (P0L , Pa) on a log–log scale. In this model (Eq. (6)) either adsorption and absorption mechanisms are effective (Pankow, 1994). 2C . P
logK P ¼ log4
3
TSP 5
mL T
p
Contents lists available at ScienceDirect
Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv
Ambient polychlorinated biphenyl levels and their evaluation in a metropolitan city S. Levent Kuzu ⁎,1, Arslan Saral 1, Gülsüm Summak 1, Hatice Çoltu 1, Selami Demir 1 Yildiz Technical University, Civil Engineering Faculty, Environmental Engineering Department, 34220, Davutpaşa-Esenler, Istanbul, Turkey
H I G H L I G H T S • PCB concentrations are found to be correlated to mixing height during summer period. • MLR, PSCF and gas/particle particle partitioning results suggested the same outcome as evaporation of PCBs from sea. • PSCF input parameters were normalized according to mixing height for the first time.
a r t i c l e
i n f o
Article history: Received 26 June 2013 Received in revised form 4 November 2013 Accepted 5 November 2013 Available online 27 November 2013 Keywords: PCB concentrations MLR PSCF Gas/particle partitioning
a b s t r a c t In this study, summer and autumn ambient PCB concentrations were investigated in metropolitan city of Istanbul. 84 congeners were targeted from di-CBs to nona-CBs on both particle and gaseous phases. Gaseous ambient concentrations were determined to be 372 ± 134 pg·m−3, while on the particle phase this value was 49 ± 17 pg·m−3, corresponding to an average of 420 pg·m−3. About one-tenth of all PCBs lay in ambient aerosols, while 90% of all comprise 2-, 3-, 4-, and 5-CBs. Measured ambient concentrations of each congener group were tested against meteorological data. The di-CB concentrations were independent of ambient temperature while northerly winds lead to an increase in their concentrations, which was an indicator of considerable contribution to di-CB concentrations from the medical waste incineration plant in Istanbul. In contrast, other congeners' concentrations were found to be correlated with southerly winds. Being an inland sea and having been contaminated, for years, by industrial discharges along the coastline, volatilization from Marmara Sea was considered as the most probable source of other congeners. PSCF analysis was run with 12-hour trajectories to locate possible local sources and check these results. Gas/particle partitioning was applied using three different models. mr and br values for log P0L model were determined as −0.23 ± 0.09 and −3.25 ± 0.38, respectively. For absorption based log Koa model, m and b values were calculated as 0.23 ± 0.08 and −4.73 ± 0.83, respectively. © 2013 Elsevier B.V. All rights reserved.
1. Introduction Of the chlorinated benzene rings, PCBs pose a great threat to human health as well as animals, plants and so forth. For this reason, these semi-volatile organic compounds (SVOCs) have attracted much interest especially in the last two decades (Cindoruk and Tasdemir, 2010). Earlier, PCBs were used as dielectric fluids, lubricants, hydraulic and heat transfer fluids until their ban in 1970s. Currently, major sources of atmospheric PCBs are listed as leakage from previously contaminated fluids, open burning, waste incineration, evaporation from contaminated sites, landfills and sludge drying beds (Breivik et al., 2002; Hsu et al.,
⁎ Corresponding author. E-mail addresses: [email protected] (S.L. Kuzu), [email protected] (A. Saral), [email protected] (G. Summak), [email protected] (H. Çoltu), [email protected] (S. Demir). 1 Tel: +90 212 383 53 78; fax: +90 212 383 53 58. 0048-9697/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.scitotenv.2013.11.031
2003). Apart from these closed or semi closed systems, some production activities and combustion processes produce PCB emissions as byproduct. In an inventory study, considering each production type which has PCB emission factor, 920 kg·yr− 1 of industrial PCB release to atmosphere is estimated within Turkey (Kuzu et al., 2013). Among the production activities, steel production by electric arc furnaces with pre-heating has the highest emission factor (Odabasi et al., 2009). Once these compounds are emitted from the source, they tend to partition at different media due to their semi-volatile characteristics (Atkinson, 1991). However, equilibrium is never reached and a continuous transfer of PCBs takes place between distinct media due to dynamic structure of the atmosphere and other compartments of the Earth. Thus, researchers inclined to understand the behavior of the compounds in the environment (Carlson and Hites, 2004). Transport of these compounds in the atmosphere is the major pathway from the point of emission until their deposition (Chen et al., 2009).
14
S.L. Kuzu et al. / Science of the Total Environment 472 (2014) 13–19
Further, these compounds reach water and soil bodies through atmospheric deposition mechanisms (Birgul and Tasdemir, 2011). Istanbul is located on both European and Asian sides, separated by the Bosphorus Strait. The city is surrounded by the Black Sea on the northern side and Marmara Sea on the southern side. Istanbul is on the route of dominant atmospheric motion through which contaminants are transported over Europe, Asia and Mediterranean region (Karaca et al., 2009). The main aims of this study is to i) investigate the ambient PCB levels in this thirteen-million-city with no previous reports on ambient levels of PCBs, ii) identify potential source regions to the ambient PCB levels, iii) determine gas/particle partitioning by present models and discuss their results. 2. Materials and methodology 2.1. Sampling Sampling was conducted within Davutpasa Campus of Yildiz Technical University in Istanbul, Turkey (41°01′26″ N and 28°53′16″ E). The campus area is surrounded by the main coach station of Istanbul on the northern side, O-1 highway on the southern side where traffic density is more than 100,000 cars per day (Onat and Stakeeva, 2013), and a great number of light and heavy industrial facilities from various sectors on western and eastern sides. A high volume air sampler (HVAS) was used to collect both particle and gaseous phase pollutants (Tecora ECHO HiVol). HVAS was equipped with a filter holder and cartridge. Ambient air was first drawn through glass fiber filter (GFF) of 10 cm in diameter and then through poly urethane foam (PUF) cartridge consisting of two identical foam layers of 5 cm in diameter and 6 cm in thickness. The foam layers were in series. The sampler was operated at 350 L·min−1 for 24 h which corresponds to approximately 500 m3 of ambient air sample. The device was calibrated before each sampling campaign. Samples were collected from May 2012 to October 2012, representing summer and autumn seasons. Davis Vantage 2 Pro online weather station was employed to obtain meteorological data during the sampling. The mixing height data were gathered from Atmospheric Research Laboratory (NOAA, 2012). Gravimetric measurements were conducted by AND GR-202 analytical microbalance. Before sampling process, GFFs were wrapped with aluminum foil and heated at 450 °C in a furnace for 6 h to remove any organic residuals. Then, the filters were conditioned in a desiccator at a temperature of 20 ± 2 °C and a relative humidity of 50 ± 5%.
Samples were then cleaned up and fractionated through a column of 3 g silicic acid (3% water), 2 g neutral alumina (6% water), and 2 g Na2SO4 (Falconer et al., 1995; Tasdemir et al., 2004), which was prewashed with 20 mL of DCM and 20 mL of PE, respectively. Next, the sample in 2 mL of hexane was poured to the column and PCBs were eluted with 20 mL of PE. The procedure to switch the solvent was repeated and the final extract was taken into hexane again. These extracts were cleaned with H2SO4 and concentrated to 1 mL by stripping with ultra pure nitrogen gas. At the last stage internal standard mixtures of PCB#30 and PCB#121 of 5 ng·ml−1 were included for volume correction purpose. Quantification of PCB concentrations was performed by a GC-ECD system (Perkin Elmer Clarus 500) equipped with a HP-5MS (30 m × 0.25 mm × 0.25 μm), capillary column. The column temperature was kept constant at 70 °C for 2 min and then raised to 150 °C at a rate of 25 °C·min−1, to 200 °C at a rate of 3 °C·min−1, and finally to 280 °C at a rate of 8 °C·min−1. The temperature was kept constant at 280 °C for 10 min. The inlet temperature 250 °C and detector temperature was 320 °C. Ultra pure helium was used as the carrier at a flowrate of 1.2 mL·min−1. The makeup gas (nitrogen) flowrate was 25 mL·min −1 . No split flow was used during the analysis. The calibration was accomplished using seven standard solutions with concentrations ranging from 0.1 ng·μL−1 to 30 ng·μL−1 After each 10 sample injections, stability was checked with the medium standard. The average value of coefficients of determination for distinct PCB congener was approximately 0.995. The investigated 84 congeners are those presented in Fig. 4 and additionally, PCB#7/9, PCB#6, PCB#12/13, PCB#18, PCB#100, PCB#66/95, PCB#92, PCB#123. 2.3. Quality assurance/quality control (QA/QC) All samples were spiked with surrogate standards and the recovery efficiencies are shown in Table 1. Least volatile congeners have higher recovery efficiencies for both PUF plugs and TSP filters. Usually PUF plugs had higher recovery efficiency among the same congener group. Limit of detection (LOD) for each PCB congener was assumed as its blank signal plus three times its standard deviation (Cindoruk and Tasdemir, 2007; Odabasi et al., 1998). Obtained LOD values for individual PCB congeners varied between 0 and 1.6 pg. Sample concentrations under LOD value were ignored. Blank samples were collected for each set of analysis. All results were blank corrected. 3. Results and discussion
2.2. Extraction and analysis
3.1. Ambient PCB concentrations
Following the sample collection, the samples were taken to the laboratory for gravimetric analysis and extraction. PCB surrogate standard mixture of PCB #14, #65, and #166 (5 ng·mL−1 for each congener) was added to the PUF cartridges and GFFs prior to the extraction procedure. The PUF cartridges were extracted with 1:1 (v/v) acetone (ACE)/ hexane (HX) mixture for 24 h in a Soxhlet extractor (Cetin et al., 2007). GFF contents were extracted in an ultrasonic bath with a mixture of dichloromethane (DCM)/petroleum ether (PE) of 1:4 (v/v) (Cindoruk and Tasdemir, 2010). Following the extraction procedure, filters were soaked in a glass jar capped with Teflon lined lids containing 25 mL of solvent mixture for 30 min. A second aliquot was formed with the same amount of solvent mixture to continue extraction for another 30 min. At the end of 1 h, the solvent mixture was evaporated and the volume was reduced to 5 mL in a rotary evaporator, which is followed by the switching of the solvent mixture to hexane by the addition of 15 mL of hexane. The latter was repeated twice. The final mixture was stripped gently by constant flow of ultra-pure nitrogen gas until its volume reached to 2 mL.
Both gaseous- and particulate-phase PCB concentrations in ambient air were measured in Istanbul between May 2012 and November 2012. Results are shown in Fig. 1 along with the atmospheric mixing heights during the sampling periods. Average gaseous phase PCB concentrations were found as 372 ± 134 pg·m−3 whereas particle phase PCB concentrations were 49 ± 17 pg·m−3. Total ambient PCB concentration was 420 pg·m−3 on average. About 12% of the total PCB
Table 1 Recoveries of the surrogate standards. Statistical parameters Recoveries of TSP filters
Recoveries of PUF plugs
PCB#14 PCB#65 PCB#166 PCB#14 PCB#65 PCB#166 Average Standard deviation 75th percentile Median 25th percentile
78.8 13.3 85.6 78.3 70.2
81.2 12.3 87.8 85 75.5
88.6 14.4 95.2 89.4 81.7
79.5 15.4 86.9 77.6 69.3
80.2 12.8 86.5 83.4 73.1
96.6 14.5 106.5 93.5 85.8
S.L. Kuzu et al. / Science of the Total Environment 472 (2014) 13–19
15
Gas
Particle
30
Percentage (%)
25 20 15 10 5 0 2-CB
3-CB
4-CB
5-CB
6-CB
7-CB
8-CB
9-CB
PCB homologues Fig. 2. Homologue PCB distribution on particle and gaseous phase.
et al., 1997; Sofuoglu et al., 2004) so far to relate the meteorological data with SVOC concentrations. Fig. 1. Ambient PCB concentrations.
concentrations lay in particulate phase. Total ambient concentrations and their distribution in particulate and gaseous phases showed good agreements with the results of a previous study in Bursa, Turkey which is located about 80 km south of Istanbul (Cindoruk and Tasdemir, 2010). Further, similar results were obtained in South Korea (Yeo et al., 2003). There hadn't been any published papers reporting PCB concentrations in Istanbul before. PCB concentration values were available for four different cities in Turkey. An average of 105 pg·m−3 PCB concentration for 14 congeners was reported in Bolu located to the Northwestern part of Turkey (Yenisoy-Karakas et al., 2012). Cetin et al. (2007) reported results between 314 and 3136 pg·m−3 for 41 congeners at an urban and industrial site, respectively, in Aliaga, İzmir. In Bursa (Cindoruk and Tasdemir, 2010), the concentrations were ranged between 316 and 570 pg·m−3 for 83 congeners. This study was conducted at four different sites representing coastal, urban, semi-rural and residential sites. Results were not differed as much as in Aliaga. The last available concentrations were measured in Konya, located in the middle of Turkey. Ozcan and Aydin (2009) observed an average concentration of 106 pg·m−3 at urban site. Regression was performed to ambient gaseous phase concentrations and mixing height values to checked whether the concentrations are correlated with mixing heights or not. However, it was discovered that results are not correlated (p b 0.15). As a next step data sets were divided into two groups as summer and fall data. Regression was again performed for each data group. Summer results were borderline statistically significant (p b 0.07), whereas fall results were statistically insignificant (p b 0.46). Atmospheric concentrations are likely to be driven by a continuous source during the summer period. It could be deduced that evaporation from previously contaminated sites such as soil or water surface could be the major PCB source of gaseous phase concentrations in warmer seasons. PCB homologue groups were illustrated for gaseous and particle phases in Fig. 2. Di-CBs to penta-CBs constitute approximately 87% of the total distribution. The distribution of di-CBs to penta-CBs for gaseous phase and particle phase was discovered to be 89% and 73%, respectively. Higher chlorinated biphenyls tend to be present in the particle phase.
3.2. Relationship between meteorological parameters and gaseous-phase concentrations In order to investigate the effect of meteorological conditions on PCB concentrations, multiple linear regression (MLR) was performed. MLR has been utilized by several researchers (Bozlaker et al., 2008; Hillery
ln P ¼ m1
. 1
T
þ m2 ðuÞ þ m3 ð cosWDÞ þ constant:
ð1Þ
In Eq. (1), the terms P, T, u and WD are corresponding to gas-phase partial pressure (atm), average ambient temperature (°K), wind speed (m·s− 1) and prevailing wind direction (degrees), respectively. m1, m2, and m3 are the regression parameters. Determined regression parameters are listed in Table 2. m1 values are proportional with the reciprocal temperature which means the higher temperature, the higher PCB concentration. m2 indicates that PCB concentrations increase when wind is blowing at slower speed. Negative m3 values are corresponding to southerly directions while positive values are corresponding northerly directions. Except di-CBs, vast majority of PCB homologue group concentrations are increased with the increasing temperature. The dependence of concentration to temperature may be attributed to the evaporation of previously deposited PCBs from terrestrial regions. Negative m2 values show that at higher wind speeds PCBs are swept by advection, consequently lower concentration is observed. The insignificance of di and tri CBs can be explained by local sources. A medical waste incineration plant is located at 20 km north of the sampling location. Jansson et al. (2011) expressed that flue gas from incineration plants are dominated by the least chlorinated PCB congeners. Moreover according to m3 value higher concentrations are observed when wind blows from the North. Finally, according to statistical results, it can be concluded that di-CBs are not related with meteorological parameters. According to m3 value, highest contribution in terms of wind direction occurs from southerly direction. Marmara sea is to the south of the sampling location where it is thought to be a polluted inland sea from the industrialized cities on the vicinity along the Marmara coastline (Gunindi and Tasdemir, 2010). Some other researchers stated that marine traffic is also accompanying to industrial SVOCs which is being transported from surrounding cities (Coelhan et al., 2006; Okay et al., 2009). Bursa is located to the south of the Marmara Sea. In this city,
Table 2 Multiple linear regression parameters.
Di-CBs Tri–CBs Tetra-CBs Penta-CBs Hexa-CBs Hepta-CBs Okta-CBs Nona-CBs ΣCBs
m1
m2
m3
r2
p1
p2
p3
303 −2259 −8528 −6505 −9522 −8628 −11,202 −7143 −4465
−0.06 −0.34 −0.26 −0.29 −0.35 −0.35 −0.32 −0.10 −0.25
0.06 −0.43 −0.90 −0.28 −0.19 −0.24 0.11 −0.35 −0.48
0.01 0.38 0.55 0.41 0.30 0.30 0.19 0.23 0.42
0.95 0.54 0.04 0.03 0.04 0.05 0.08 0.12 0.16
0.78 0.03 0.11 0.02 0.05 0.05 0.20 0.55 0.06
0.84 0.06 b0.01 0.11 0.45 0.34 0.76 0.19 0.02
16
S.L. Kuzu et al. / Science of the Total Environment 472 (2014) 13–19
Cindoruk and Tasdemir (2010) stated that PCBs are likely to be transported from Marmara Sea according to MLR results. They concluded that evaporation from sea surface and air–water exchange could be an active mechanism. Considering our results, this fact can be confirmed and it is observed that Marmara sea is acting as a PCB source. 3.3. PSCF analysis PSCF analysis was executed for a further evaluation of the previous outcomes of MLR. PSCF can be briefly described as conditional possibility, characterizing the spatial distribution of possible geophysical source locations made out by using trajectories reaching to the sampling site (Wimolwattanapun et al.). Via this method, probability field is developed which equates excessive concentration values at the receptor point with their probable source locations (Uygur et al., 2010). In the theory of the model, it is supposed that the end point of a trajectory in a grid cell is assumed to gather the pollutant in concern. After that, the incorporated pollutant is supposed to be transported to the sampling point along the trajectory pathway with the air parcel (Heo et al., 2009). PSCF can be calculated using Eq. (2): PSC F i; j ¼
M i; j Ni; j
ð2Þ
where ΣNi,j is the total number of trajectory segment endpoints in the ijth grid cell during the whole sampling period and ΣMi,j is the total number of trajectory segment endpoints in the same grid cell which is corresponding to samples that are exceeding the threshold criteria. Small ni,j values cause elevated PSCF results with high uncertainties. The small values usually encountered at more distant grids from the source due to sparse distribution of the trajectories. To reduce the effect of small ni,j values PSCF result multiplied by an arbitrary weighting function Wi,j (Polissar et al., 1999). Assigned Wi,j values are listed in Eq. (3).
W i; j ¼
8 > > < > > :
1:00 80bni; j 0:70 20bni; j ≤80 0:42 10bni; j ≤20 0:05 ni; j ≤10
:
ð3Þ
It was told previously that, in the present study PSCF analysis was executed to support the MLR results. According to MLR results two sources are likely to contribute ambient PCB concentrations i) Marmara Sea to the south and ii) medical waste incineration plant to the north. For that reason back-trajectories were calculated for 12 h. This selection
leads us to resolve short-range sources. The grid cells are formed with a resolution of 0.5° × 0.5° between longitudes of 25 E and 35 E and latitudes of 40 N and 44 N. A threshold criteria need to be chosen to attain some of the trajectories as polluted. Conventionally, average value of pollutant concentrations is regarded as the threshold. However, we think that, this doesn't reflect the real case due to dynamic structure of the atmosphere. A situation of high concentration and low mixing height doesn't mean that the corresponding trajectory at the time of sampling period induced the excessive pollutant amount. In contrast, any case of concentration value that is slightly less than threshold value with high mixing height doesn't mean that this trajectory should not be selected as polluted. Therefore, we decided to choose the threshold criteria as normalized ambient PCB concentrations according to atmospheric planetary boundary layer values. For the calculation, PCB concentrations are multiplied by mixing heights reported for the corresponding sampling dates. PCB loads per unit area were determined as a result of this normalization process. Trajectories were selected considering the mean value and the criterion value was set to the 75th percentile. Results are depicted in Fig. 3. TrajStat v.1.2.2.6 was used for calculation and mapping purpose (Wang et al., 2009). Considering the factors, two prevailing regions are likely to affect PCB concentrations in Istanbul. One of them is the Marmara Sea as stated before while the other one seems as Black Sea at the North. During the warmer seasons volatilization from water surfaces can be regarded as an important PCB source (Agrell et al., 1999). Even, there is a concern on the occurrence of PCBs at the marine environment (Li et al., 2012). Also, it should be pointed out that northerly trajectories reach to the sampling point passing over the medical waste incineration plant. As a result, MLR results seem to be in accordance with the PSCF results. 3.4. Gas/particle partitioning Particle and gas partitioning are crucial in determining the fate of chemicals in the atmosphere (Hoff et al., 1996; Yeo et al., 2003). Several models have been developed so far to predict the gas/particle partitioning. One of these models is the Junge–Pankow adsorption model. This model was developed according to linear Langmuir isotherm (Tasdemir et al., 2004). In the model it is supposed that SVOC compounds are adsorbed to active sites on the surface of the particle (Harner and Bidleman, 1998). Junge–Pankow model is shown in Eq. (4). ∅¼
cθ p0L þ cθ
Fig. 3. PSCF results during the sampling period.
ð4Þ
S.L. Kuzu et al. / Science of the Total Environment 472 (2014) 13–19
∅¼
ð5Þ
Ratio of measured and predicted values according to Junge–Pankow model were plotted in Fig. 4. For low molecular weighted (MW) congeners, measured values are higher than the predicted values. The model underestimated MW congeners. However an opposite case is present for higher MW congeners. In that case, model results over estimate the partitioning fraction. Observed results are consistent with the previous studies in Chicago and Bursa atmosphere (Tasdemir et al., 2004; Cindoruk and Tasdemir, 2007). Lee and Jones (1999) stated that, c and θ values, which are assumed, are effective in achieving over and under estimation. In addition, black carbon particles affect gas/particle partitioning (Lohmann and Lammel, 2004). These contributions are thought to be effective in deviating the values from 1 in Junge–Pankow adsorption model. Another model was applied in which gas/particle partitioning coefficient (Kp, m−3·μg) is plotted against supercooled vapor pressure (P0L , Pa) on a log–log scale. In this model (Eq. (6)) either adsorption and absorption mechanisms are effective (Pankow, 1994). 2C . P
logK P ¼ log4
3
TSP 5
mL T
p
