Environ. Sci. Technol. 2004, 38, 3279-3285
Quantitative Source Identification of Dioxin-like PCBs in Yokohama, Japan, by Temperature Dependence of Their Atmospheric Concentrations I S A M U O G U R A , * ,† SHIGEKI MASUNAGA,‡ AND J U N K O N A K A N I S H I †,‡ Research Center for Chemical Risk Management, National Institute of Advanced Industrial Science and Technology, 16-1 Onogawa, Tsukuba 305-8569, Japan, and Graduate School of Environment and Information Sciences, Yokohama National University, 79-7 Tokiwadai, Hodogaya-ku, Yokohama-shi, Kanagawa 240-8501, Japan
The source and environmental behavior of dioxin-like polychlorinated biphenyls (PCBs) together with other PCBs and polychlorinated dibenzo-p-dioxins and dibenzofurans (PCDDs/PCDFs) were evaluated based on analysis of variations in their atmospheric concentrations in Yokohama, Japan. Potential factors responsible for variations in the atmospheric concentrations of the congeners were investigated by principal component analysis and multiple regression analysis of the data. Two major variations were seen: one had strong temperature dependence, while the other had no significant temperature dependence. A possible explanation for this difference is that the former is related to congeners released by volatilization (e.g., volatilization from commercial PCB products and past polluted environments), while the latter is related to congeners emitted from thermal processes. The relative contributions of dioxin-like PCBs released by volatilization and those emitted from thermal processes were estimated based on the temperature dependence of the atmospheric concentrations. The results suggest that both dioxin-like PCBs emitted from thermal processes and those released by volatilization are significant sources of air pollution in this area in terms of the toxic equivalent (TEQ) for dioxin-like PCBs. We demonstrated that the present approach based on variations in atmospheric concentrations can be useful in providing a qualitative as well as quantitative understanding of source information.
Introduction Twelve polychlorinated biphenyls (PCB-77, -81, -105, -114, -118, -123, -126, -156, -157, -167, -169, and -189) (IUPAC nos.), referred to as dioxin-like PCBs in terms of their toxicity, have been evaluated together with polychlorinated dibenzop-dioxins and dibenzofurans (PCDDs/PCDFs) (1). In Japan, the contribution of dioxin-like PCBs to the toxic equivalent (TEQ) of human intake is generally higher than that of PCDDs/PCDFs (2). Understanding the source contributions * Corresponding author phone: +81 29 861 8907; fax: +81 29 861 8415; e-mail:
[email protected]. † National Institute of Advanced Industrial Science and Technology. ‡ Yokohama National University. 10.1021/es0354622 CCC: $27.50 Published on Web 05/08/2004
2004 American Chemical Society
of dioxin-like PCBs in the environment is essential to developing effective countermeasures against their pollution. One of the well-known potential sources of dioxin-like PCBs in Japan is the emissions from commercial PCB products. In Japan, commercial PCB production began in 1954. The annual PCB usage increased from 1954 (200 tons yr-1) to 1970 (10 120 tons yr-1), and the production and new use of PCBs were both phased out in 1972 (3). During the period when PCBs were on the market, very high levels of environmental PCB contamination were found in the vicinity of plants producing PCBs and PCB-containing items, factories where PCBs were used, and recycling paper mills (3). Because it has until quite recently been difficult to establish a consensus among local residents regarding the siting of PCB treatment facilities, waste PCB products and PCB-containing items have been kept in depositories for several decades. It is suspected that the emission of PCBs into the environment by volatilization from waste PCB products and PCB-containing items continues today (4). Moreover, the volatilization of PCBs from past polluted environments is also expected to be occurring. However, the quantity of dioxin-like PCBs as well as other PCBs entering into the environment by volatilization is not yet known. Another major potential source of dioxin-like PCBs is the emissions from thermal processes. It is known that PCBs and PCDDs/PCDFs are formed as byproducts in certain thermal processes. In particular, it has been reported that non-ortho-PCBs are formed during the incineration of municipal waste (5-7). As of 2001, there were approximately 17 000 waste-incineration facilities, including municipal waste incinerators, industrial waste incinerators, and smallscale waste incinerators, and approximately 1000 industrial thermal plant facilities targeted for regulation of dioxin (PCDD/PCDF and dioxin-like PCB) emissions, in Japan (8). According to the Japanese dioxin emission inventory study, the amount of emissions for TEQ of dioxins into the air was estimated to be approximately 8000 g yr-1 in 1997 and then decreased to approximately 1000 g yr-1 in 2002 after enforcement of the regulation of dioxin emissions began in 1997 (9). Because these estimates were based on only a limited number of collected samples, and as it remains uncertain what percentage of the dioxin-like PCBs in the environment is derived from known sources, it is important to evaluate source contributions to the amounts of dioxin-like PCBs in the environment using actual environmental data. In the present study, we analyzed variations in the atmospheric concentrations of dioxin-like PCBs together with other PCBs and PCDDs/PCDFs in Yokohama, Japan, to evaluate the source contributions to the atmospheric concentrations of these congeners. Potential factors responsible for variations in the atmospheric concentrations of the congeners were investigated by principal component analysis and multiple regression analysis of the data. The relative contributions of dioxin-like PCBs released by volatilization and those emitted from thermal processes to the atmospheric concentrations were estimated based on the temperature dependence of the atmospheric concentrations.
Potential Processes Potential processes that can affect the atmospheric concentrations of congeners are shown in Figure 1. Congeners Released by Volatilization. With regard to volatilization of a congener from organic matter (e.g., PCB mixtures, PCB-containing oil, or organic matter in a polluted environment), if the concentrations of the congener in the gas-phase close to the organic matter Cg (pg m-3) and in the VOL. 38, NO. 12, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3279
obtained by using 1/H (unitless Henry’s law constant) instead of KOA in the same manner as the volatilization of a congener from an organic matter phase. Then, the slope of the log Cvol vs 1/T relationship should be the slope of the log H vs 1/T relationship. The slopes of the plot of log H vs 1/T for tetrato heptachlorobiphenyls have been found to range from -3900 to -3200 (16-18). Congeners Emitted from Thermal Processes. Assuming that there is little difference between the elimination rate of a congener released by volatilization and that of the congener emitted from thermal processes, the atmospheric concentration of the congener emitted from thermal processes Cincine (pg m-3) under a steady-state condition can be expressed by
E ) k × Cincine FIGURE 1. Atmospheric PCBs released by volatilization and emitted from thermal processes. organic matter Cs (pg m-3) are in equilibrium, then
Cs/Cg ≈ a × KOA
(1)
where a (unitless) is the factor for conversion from the octanol-air partition coefficient to the organic matter-air partition coefficient, and KOA (unitless) is the octanol-air partition coefficient. The log KOA has an almost linear relationship to the reciprocal temperature (10)
log KOA ≈ A/T + B
(2)
where A and B are constants in the range of the ambient temperature of interest and T (K) is the ambient temperature. The slopes (A) of the plot of log KOA vs -1/T for tetra- to heptachlorobiphenyls have been found to range from -4900 to -3800 (10-12). According to eqs 1 and 2,
log(Cs/Cg) ≈ A/T + B + log a
(3)
When the volatilized congener is transported to ambient air, the atmospheric concentration of the congener released by volatilization Cvol (pg m-3) under a steady-state condition can be expressed by
F × Cg ) k × Cvol
(4)
where F (s-1) is the rate of transport from the source to ambient air and k (s-1) is the first-order rate constant of elimination from ambient air by deposition, degradation, advection, and so forth. When a volatilized congener is transported to ambient air and partly sorbed into the suspended particles, Cvol should be the total atmospheric concentration in both gas and particle phases rather than the atmospheric concentration in the gas phase. According to eqs 3 and 4,
log Cvol ≈ - A/T - B - log a + log Cs + log F - log k (5) Assuming that Cs is so large that it does not change due to temperature-driven volatilization and that a, F, and k are independent of temperature, the plot of log Cvol vs 1/T should exhibit a linear relationship with a slope of -A. The temperature dependences of the atmospheric concentrations of PCBs as well as those of semivolatile organic compounds have been reported by a number of studies (13-15). Furthermore, with regard to volatilization of a congener from polluted water (e.g., water in a polluted sea, river, or lake), the relationship between 1/T and log Cvol can be 3280
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 12, 2004
(6)
where E (pg m-3 s-1) is the emission rate for a congener from thermal processes. Normalization by a Reference Compound. Because the atmospheric concentration of a congener is variable with changes in k according to meteorological conditions, the atmospheric concentration of the congener of interest can be normalized by dividing it by the atmospheric concentration of a reference compound Cref (pg m-3). Assuming that the reference compound is a compound emitted from thermal processes and that the elimination rate of the reference compound is equal to that of the congener of interest
Eref ) k × Cref
(7)
where Eref (pg m-3 s-1) is the emission rate for the reference compound from thermal processes. According to eqs 5-7,
log(Cvol/Cref) ≈ - A/T + B′ - log Eref
(8)
log(Cincine/Cref) ) log(E/Eref)
(9)
It is assumed that both E and Eref are independent of temperature and that E/Eref is constant. If the congener of interest is mostly released by volatilization, the plot of log(Cair/Cref) vs 1/T should exhibit a linear relationship with a slope of -A, as in eq 8 (where Cair (pg m-3) is the atmospheric concentration of the congener of interest). In contrast, if the congener of interest is mostly emitted from thermal processes, log(Cair/Cref) should be constant against the temperature change as in eq 9. From this, the relative contributions of the congener released by volatilization and the congener emitted from thermal processes to the total atmospheric concentration of the congener can be determined. The magnitude of the slope can be considered as an indication of the relative contribution of the congener released by volatilization. This approach is valid in urban or industrial areas where the contribution of the advection of background air into the area is negligible. If the congener of interest is derived primarily from the advection of background air into the area, log(Cair/Cref) may be almost constant against temperature change. The present approach partly follows the concept discussed by Wania et al. (13). They have introduced an approach to assess the relative magnitude of evaporation from sources close to the sampling locations vs advection of background air into the area based on the temperature dependence of atmospheric concentrations of semivolatile organic compounds.
Materials and Methods Sample Collection. Sampling was conducted on the roof of a building at Yokohama National University in Yokohama,
which is a bustling city in an urban industrial area with a population of over 3 million people and is Japan’s second largest city after Metropolitan Tokyo. Air samples were collected by a high-volume air sampler (DHV-1000S, Shibata Scientific Technol. Ltd.) equipped with a quartz fiber filter (QFF) to collect compounds in the particle phase and two polyurethane foam plugs (PUFs) to collect compounds in the gas phase. Before sampling, QFFs were heated to approximately 600 °C, and PUFs were cleaned with acetone using a Soxhlet extractor for 8 h to reduce contamination. Each sample was collected for 48 h at a flow rate of 0.566 m3 min-1. The samples were collected approximately twice a month during the period from Oct 1998 to Mar 2000 (n ) 22). Sample Analysis. The method of analysis basically followed that of our previous study (19). The QFFs and the PUFs were separately analyzed. After a mixture of 13C12labeled internal standards (12 dioxin-like PCBs, PCB-170 and -180 (IUPAC nos.), and 16 2,3,7,8-substituted PCDDs/PCDFs) was added, the QFFs were extracted with toluene using a Soxhlet/Dean-Stark extractor for 16 h, and the PUFs were extracted with acetone using a Soxhlet extractor for 16 h. The extracts were cleaned by sulfuric acid/silica gel column chromatography (or treated with concentrated sulfuric acid and cleaned by silica gel column chromatography). The extracts were divided into a fraction of non-ortho dioxin-like PCBs and PCDDs/PCDFs and a fraction of mono-ortho dioxin-like PCBs and other PCBs using activated-carbonimpregnated silica gel column chromatography. These final extracts were concentrated to 25 µL and were analyzed using HRGC/HRMS (HP6890, Hewlett-Packard, and Autospec-Ultima, Micromass). The mass spectrometer was operated in the EI mode at a resolution of R > 10 000 (10% valley). Tetra- to octa-chlorinated congeners of PCBs and PCDDs/PCDFs were quantified using a DB-5 column (60 m, J&W Scientific). 2,3,7,8-Substituted PCDDs/PCDFs not separable by a DB-5 column were quantified using a DB-17 column (60 m, J&W Scientific). Peaks were assigned according to the literature (20-23). Toxic equivalency factors (TEFs) of WHO (1998) were used for the quantification of TEQ. Quality Control and Quality Assurance. Average recoveries of 13C12-labeled internal standards for the non-ortho PCBs, the mono- and di-ortho PCBs, and the 2,3,7,8substituted PCDDs/PCDFs were 60-74%, 68-91%, and 5884%, respectively. Procedural blanks consisting of clean and unused QFF and/or PUF, extracted and prepared using an procedure identical to that used for the samples, were analyzed in parallel with the samples. The quantification limit for each congener was determined as five times the maximum blank value in the blank samples. The quantification limits for congeners of which values could not be detected in the blank samples were considered equal to the determination limits, which were calculated using a peak-to-peak noise ratio of 10. For the results and discussion in this paper, the total (gas and particulate) concentrations were used. When the concentration in either gas or particle phase for each congener for each sample was below the quantification limit, the total concentration was calculated only if the quantified concentration of one phase was five times greater than the quantification limit for the other phase. The concentrations below the quantification limit were regarded as zero. Breakthrough from the first to the second PUF was investigated at volumes of approximately 1700 m3 and ambient temperatures of 14-23 °C, and it was found that the amounts of congeners on the second PUF were at the same level as those in the blank samples.
Statistical Analysis Principal Component Analysis. Principal component analysis of the data was conducted in order to detect similarities, differences, and relationships of the variations in the atmospheric concentrations among congeners and to evaluate the number of significant factors affecting variations in the atmospheric concentrations. Forty-three variables for PCB congeners and 82 for PCDD/PCDF congeners corresponding to the chromatographic peaks on DB-5 that could be quantified in more than 15 of 21 samples were subjected to the principal component analysis. The atmospheric concentrations were transformed into logarithms of the atmospheric concentrations, since the atmospheric concentrations varied widely among samples and since it was expected that the logarithm of the atmospheric concentration of one congener has a linear relationship to that of another congener when both are released by volatilization, as in eq 5. The variances of the logarithms of the atmospheric concentrations of PCBs and PCDDs/PCDFs were normalized to unity using the correlation matrix. The correlation matrix was calculated with pairwise deletion of missing data. The eigenvectors were normal-varimax-rotated for better interpretation of the results. Multiple Regression Analysis. Multiple regression analysis of the data was conducted to analyze the factors affecting variations in the atmospheric concentrations of congeners. Ambient temperature, wind speed, and sampling date were selected as potential factors, since they could be obtained easily and since they are considered to affect the atmospheric concentrations of congeners. Reciprocal temperatures, the logarithms of wind speeds, and sampling dates were used as independent variables, and the logarithms of the atmospheric concentrations of PCBs and PCDDs/PCDFs were used as dependent variables, according to the following considerations: 1. Ambient temperature: If a congener is released into the air by volatilization, it is expected that the logarithms of the atmospheric concentrations of the congener have a linear relationship to the reciprocal of the temperature, as in eq 5. 2. Wind speed: If the dilution rate for the atmospheric concentrations of congeners is correlated to wind speed, the atmospheric concentrations of congeners are considered to be correlated to reciprocal wind speed. That is, the logarithms of the atmospheric concentrations of congeners are considered to have a linear relationship to the logarithms of wind speeds. 3. Time trend: The literature (24, 25) indicates that the atmospheric concentrations of total dioxin-like PCBs and total PCDDs/PCDFs in Tokyo, Japan, have been almost exponentially decreasing from 1980 to 2001. The same tendency has been shown in old rice straws (26, 27) and in the sediment core from Lake Haruna (28), both of which are considered to have received only atmospheric inputs. Assuming that the atmospheric concentrations have been exponentially decreasing, the logarithms of the atmospheric concentrations of congeners would have a linear relationship to the sampling date. The average temperature and average wind speed during each sampling were used as variables regarding temperature and wind speed. Principal component analysis and multiple regression analysis were performed using Statistica for Windows 5.0J (StatSoft, Inc.).
Results and Discussion Atmospheric Concentrations of Dioxin-like PCBs and PCDDs/PCDFs. The observed atmospheric concentrations ranged from 0.0027 to 0.015 (mean 0.0082) pg m-3 for TEQ of dioxin-like PCBs and from 0.016 to 0.60 (mean 0.23) pg m-3 for TEQ of PCDDs/PCDFs. These concentrations were VOL. 38, NO. 12, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3281
TABLE 1. Contributions of Temperature, Wind Speed, and Sampling Date to Variations in the Atmospheric Concentrations of PCBs and PCDDs/PCDFs standardized partial regression coefficients congener
FIGURE 2. Plot of loadings on PC-1 and PC-2 for the logarithms of the atmospheric concentrations of PCDDs/PCDFs and PCBs. The numbers indicate IUPAC nos. of PCBs. close to the average for urban areas in Japan (0.012 pg m-3 for TEQ of dioxin-like PCBs and 0.22 pg m-3 for TEQ of PCDDs/PCDFs) and were higher than those in the background area (0.0018 pg m-3 for TEQ of dioxin-like PCBs and 0.015 pg m-3 for TEQ of PCDDs/PCDFs) (29). These results suggest that the contribution of advection of background air into the area to the atmospheric concentrations of dioxinlike PCBs and PCDDs/PCDFs is small. Relationships among Variations in the Atmospheric Concentrations of Congeners. We conducted principal component analysis of the data. The results indicated that the first two principal components, PC-1 and PC-2, accounted for 66 and 23% of the total variance, respectively. Because the third principal component accounted for less than 5% of the total variance, the number of significant factors was considered to be two. The plot of loadings on PC-1 and PC-2 is shown in Figure 2. The congeners related to PC-1 that had high loading included the majority of PCDDs/PCDFs and PCB-126, -169, and -189. These congeners were located close to each other, indicating that they are highly correlated to one another. These results imply that the major emission source for these congeners is identical and that they have similar environmental behaviors. On the other hand, the majority of PCB congeners were related to PC-2, indicating that the variations in their atmospheric concentrations are different from the variations in the atmospheric concentrations of the congeners related to PC-1. These results suggest that the congeners related to PC-2 are different from the congeners related to PC-1 in terms of emission sources or environmental behaviors. The congeners such as PCB-81, -156, and -157 located near the radius of the region between PC-1 and PC-2 in the plot of loadings are considered to be influenced by both factors. Factors Affecting the Atmospheric Concentrations of Congeners. We performed multiple regression analysis of the data. The multicollinearity among the independent variables was found to be insignificant by means of variance inflation factors (VIFs). That is, the values of VIFs were between 1.5 and 1.8 and satisfied a criterion of potential multicollinearity problems (i.e., VIF ) 10) (30). The results for selected PCBs are shown in Table 1 together with those for 2,3,4,7,8-pentachlorodibenzofuran (2,3,4,7,8penta-CDF) and homologue totals of PCDDs/PCDFs (TetraCDD, Penta-CDD, Hexa-CDD, Hepta-CDD, Octa-CDD, TetraCDF, Penta-CDF, Hexa-CDF, Hepta-CDF, and Octa-CDF) as representatives of the PCDD/PCDF congeners. The magnitudes of the absolute values of standardized partial regression 3282
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 12, 2004
PCB-151 PCB-123 PCB-124 PCB-174 PCB-77 PCB-81 PCB-191 PCB-157 PCB-189 PCB-126 PCB-169 2,3,4,7,8-PentaCDF Tetra-CDD Penta-CDD Hexa-CDD Hepta-CDD Octa-CDD Tetra-CDF Penta-CDF Hexa-CDF Hepta-CDF Octa-CDF a
no. of samples
1/temp
multiple log wind sampling correlation speed date (R2)
22 18 22 22 20 22 21 18 21 22 20 22
-1.15a -1.28a -1.20a -1.10a -0.98a -0.83a -0.33 -0.21 0.17 -0.14 0.04 0.20
-0.31a -0.52a -0.46a -0.34a -0.56 -0.64a -0.77a -0.60a -0.58a -0.73a -0.53a -0.48a
-0.34a -0.45 -0.54a -0.43a -0.62 -0.89a -0.57a -0.56 -0.41a -0.79a -0.82a -0.65a
0.94 0.90 0.93 0.88 0.56 0.79 0.67 0.65 0.78 0.85 0.81 0.85
22 22 22 22 22 22 22 22 22 22
0.04 0.29 0.12 -0.03 -0.14 0.07 0.12 0.07 0.005 -0.06
-0.63a -0.58a -0.63a -0.65a -0.69a -0.68a -0.65a -0.59a -0.60a -0.60a
-0.70a -0.50a -0.61a -0.70a -0.77a -0.65a -0.60a -0.60a -0.59a -0.64a
0.84 0.89 0.85 0.81 0.81 0.86 0.85 0.78 0.73 0.73
p
