Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2016 Dec 1.
Published in final edited form as: Atmos Environ (1994). 2015 Dec 1;122:791–798. doi: 10.1016/j.atmosenv.2015.10.040

Atmospheric dispersion of PCB from a contaminated Lake Michigan harbor

Andres Martinez a,*, Scott N Spak a,b, Nicholas T Petrich a, Dingfei Hu a, Gregory R Carmichael c, Keri C Hornbuckle a
PMCID: PMC4649934  NIHMSID: NIHMS735453  PMID: 26594127

Abstract

Indiana Harbor and Ship Canal (IHSC) in East Chicago is an industrial waterway on Lake Michigan and a source of PCBs to Lake Michigan and the overlying air. We hypothesized that IHSC is an important source of airborne PCBs to surrounding communities. We used AERMOD to model hourly PCB concentrations, utilizing emission fluxes from a prior study and hourly meteorology provided by the State of Indiana. We also assessed dispersion using hourly observed meteorology from a local airport and high resolution profiles simulated by the Weather Research and Forecasting model. We found that emissions from IHSC waters contributed about 15% of the observed ΣPCB concentrations close to IHSC when compared on an hourly basis and about 10% of observed annual concentrations at a nearby school. Concentrations at the school due to emissions from IHSC ranged from 0 to 18,000 pg m−3, up to 20 times higher than observed background levels, with an annual geometric mean (GSD) of 19 (31) pg m−3. Our findings indicate that IHSC is an important source of PCBs to East Chicago, but not the only source. Four observed enriched PCB3 samples suggest a nearby non-Aroclor source.

Keywords: Polychlorinated Biphenyls (PCBs), AERMOD, Weather Research and Forecasting Model (WRF), Indiana Harbor and Ship Canal (IHSC), 4-monochlorobiphenyl (PCB3)

Graphical Abstract

graphic file with name nihms735453u1.jpg

1. Introduction

Indiana Harbor and Ship Canal (IHSC) was designated as an Area of Concern by the International Joint Commission due to the presence of high levels of heavy metals, PAHs and PCBs (International Joint Commission, 2003). In 2012, the U.S. Army Corps of Engineers began a navigational dredging project where the sediments are disposed in a confined disposal facility (CDF) located next to IHSC in East Chicago, Indiana. The surrounding community is densely populated (813 people per km2, U.S. Census 2010), and there is concern that dredging will increase exposure to airborne PCBs.

The World Health Organization classifies polychlorinated biphenyls (PCBs) as semi-volatile organic compounds (SVOCs) due to their boiling point range and to differentiate from volatile organic compounds (VOCs) (World Health Organization, 1989). PCBs volatilize from surfaces and materials, such as contaminated soils, waters, and building materials (Chiarenzelli et al., 1997; Harner et al., 1995; Hu and Hornbuckle, 2010; Totten et al., 2003). Contaminated water, particularly waters that are in contact with contaminated sediments, can be a large source of gas-phase PCBs. Connolly et al. estimated that between 1978 and 1998, 1400 kg yr−1 of total PCBs (ΣPCBs) were emitted from the water of the Upper Hudson River to the atmosphere (Connolly et al., 2000). Achman et al. estimated that contaminated water of Green Bay, WI, emitted ca. 6 kg of ΣPCBs to the atmosphere during 14 days in 1989 (Achman et al., 1993). In a previous study, we estimated that 7 kg yr−1 of ΣPCBs were emitted from the contaminated waters of IHSC to the overlying air (Martinez et al., 2010). However, no prior studies have utilized those emission predictions to quantify the impact of these sources on local or regional airborne PCB concentrations.

An atmospheric dispersion model driven by local meteorological data is required to simulate local transport of SVOC emissions and resultant concentrations. AERMOD, developed by the American Meteorological Society and U.S. EPA, represents the current state of the science in regulatory atmospheric dispersion modeling (Cimorelli et al., 2005). AERMOD was developed primarily for regulatory assessments of primary criteria pollutants, air toxics, and VOCs. It is not commonly used for modeling SVOCs. Morselli et al. (2012) provides one of the few reports of AERMOD being used to predict SVOCs. They coupled a multimedia fate model with AERMOD driven by observed meteorological data to model the air concentration of two polycyclic aromatic hydrocarbons (PAHs) released from an incinerator-like point source. The primary limitation for PCB atmospheric dispersion modeling has been the lack of quantitative emissions data.

Here we report the first study to use AERMOD to predict air PCB concentrations from PCB emissions from a contaminated waterway. We used AERMOD to predict PCB congener concentrations in air directly over the water and into the air of the nearby community. To determine the relative importance of IHSC compared to other potential and unknown sources of airborne PCBs in the area, we used data from air samples collected over the water, on land near the canal, and at a nearby school. Here we report our findings with respect to: (i) transport of airborne PCB congeners; (ii) PCB congener contribution from IHSC emissions to the local atmosphere; and (iii) possible sources of PCB congeners in East Chicago and the surrounding area.

2. Methods and materials

2.1. Atmospheric dispersion model

We used Breeze AERMOD-ISC Version 7.6 to model hourly airborne PCB concentrations, running the regulatory configuration of AERMOD Version 12060 (February 29th 2012). AERMOD is a steady state dispersion plume model that includes the effects on dispersion from vertical variations in the planetary boundary layer (PBL). Under stable conditions the concentration distribution is Gaussian, both vertically and horizontally, as is the horizontal distribution in convective conditions. During convective conditions the vertical concentration distribution is described with a bi-Gaussian probability density function (Cimorelli et al., 2005).

We chose to use AERMOD because its performance in 17 field studies was superior to other models tested (Perry et al., 2005) and because it is the most widely applied regulatory atmospheric dispersion model approved by U.S. EPA (USEPA, 2014). Further, we consider the use of a steady-state model quite reasonable for such a local study, where even a 0.5 m s−1 wind will carry all fresh emissions out of the modeling domain within a one hour time step. Performance by plume and puff models has been found commensurate at these small distances (Moore et al., 2012).

We applied AERMOD at 50 × 50 m2 resolution to a receptor grid (10,000 × 12,000 m2, 48,000 cells) completely covering IHSC and the City of East Chicago, centered on Latitude 41°39′13.64″N, Longitude 87°27′15.42″W. We used 10 m resolution terrain data from the US Geological Service National Elevation Dataset (Gesch et al., 2002) processed by AERMAP (USEPA, 2004a). All receptors were placed 1 m above the surface, corresponding to the inlet height of air samplers.

PCBs were treated as conservative compounds for such local dispersion due to the relative long half-life in air, ca. 6 yr (Basu et al., n.d.; Hillery et al., 1997). Our model did not include any estimated PCB background concentrations because the temporal trends in background concentrations at East Chicago remain unknown, and we sought to understand the impact of IHSC emissions alone. Wet and dry depositions were not considered, again due to the short residence time in the small local domain.

Hourly airborne ΣPCB concentrations were modeled to support the investigation of diel patterns and the effects of hourly meteorological conditions on sampled concentrations from high volume air sampler (Hi-Vol) integrated for 10–12 h. The gas phase concentration for individual congeners was estimated from the product of the modeled ΣPCB concentration and the PCB congener fraction from the emissions.

2.2. Meteorological data

AERMOD requires time-resolved surface meteorological data and vertically resolved meteorological data, typically taken from hourly surface measurements at weather stations and profiles estimated from available upper air soundings. IHSC is located on the southwest shore of Lake Michigan, where local meteorology is strongly affected by the lake, with consistent impacts on wind speed and direction from the lake breeze circulation, the lake’s annual temperature cycle, and mixing layer height, with the marine boundary layer shallow in summer and deep in winter. The Indiana Department of Environmental Management (IDEM) distributes quality-controlled meteorological data formatted used in AERMOD for regulatory permitting applications in 7 regions of the state, using Automated Weather Observing System (AWOS) surface observations and twice-daily upper air profiles from the nearest National Weather Service radiosonde station (Indiana Department of Environmental Management, n.d.). However, the observed surface data provided and recommended by the agency come from South Bend, IN, which is located ca. 50 km east of Lake Michigan and 120 km from IHSC, and the vertical profile from Lincoln, IL, ca. 270 km southwest of IHSC on an open prairie. While these are the nearest approved stations, they might not represent local meteorological conditions on the southwest shore of Lake Michigan. Therefore, we applied and evaluated the following meteorological data sets: (i) observations provided by IDEM, (ii) hourly values simulated using the Weather Research and Forecasting Model (WRF) version 3.3, (National Center for Atmospheric Research, 2012; Sinkkonen and Paasivirta, 2000) and (iii) observed surface observations from Gary Airport, IN, and vertical values from Lincoln, IL (NOAA, National Oceanic and Atmospheric Administration) processed by AERSURFACE (USEPA, 2008) using local surface roughness data estimated from the National Land Cover Dataset 2006 (Fry et al., 2006).

To comprehensively quantify the role of observed and simulated surface meteorology, and estimated boundary layer height, we constructed and evaluated hybrid inputs representing all combinations of these three inputs, including WRF surface and vertical profiles with all wind directions adjusted to remove the hourly bias relative to Gary airport observations at 10 m, with wind speed adjusted to (iv) subtract bias and (v) scaled relative to surface ratio between observations and model; Gary airport observed values combined with (vi) debiased and (vii) scaled WRF vertical profiles; and (viii) observed surface values at Gary with the vertical profile from Lincoln. These eight meteorological data sets were input into AERMOD and the modeled concentrations were compared against PCB measurements carried out above IHSC for August 2006.

WRF simulated hourly surface and vertically-resolved meteorology over IHSC at 1.33 km horizontal resolution with 35 vertical layers, as described by Petrich et al. (2013). WRF profiles were processed using MCIP2AERMOD (Davis et al., 2008), which calculated mechanical mixing height, relative humidity, Bowen ratio, albedo, and lapse rate above mixing height; all other AERMOD input variables were taken directly from WRF. WRF was run continuously for the month of August 2006 and in four quarterly continuous simulations for 2008 (Petrich et al., 2013). Descriptive statistics of the eight meteorological data for August 2006 are presented in Table S2.

2.3. PCB emissions

Hourly ΣPCB gas emissions from the waters of IHSC were calculated as the sum of the products of the individual PCB congener air–water mass transfer coefficients, the individual PCB water concentrations and the surface area of IHSC. Air-water mass transfer coefficients were calculated as a function of the individual PCB congener physical–chemical properties, and hourly local environmental factors, including air and water temperatures, atmospheric pressure and wind speed at 10 m above the surface. This approach has been described previously and calculation details for this study are presented in the Supplementary Data (Martinez et al., 2010). We previously reported the net flux of PCB congeners, considering both emission and gas deposition. For this study, we used only the gross emission fluxes. The environmental factors were obtained from the National Data Buoy Center (NOAA). Hourly PCB water concentrations were estimated from our field measurements during August 2006 (Fig. S1). These concentrations are the only individual PCB congener water concentrations available for this system. We decided to use these values for the entire simulation, instead of speculating seasonality in concentrations.

We modeled fluxes from each individual location (n = 10) were the water samples were collected and no spatial variability was found. The water concentrations of dissolved PCBs in IHSC are mainly constant (Relative standard deviation, RSD = 45%) and are contaminated by Aroclor 1248 (Martinez et al., 2010). Thus, we used the average of all the water samples to estimate the emissions from IHSC. Modeled hourly gas emissions of ΣPCB for August 2006 ranged from 0.44 to 0.47 g h−1, with a small decrease in emission rates toward the end of month (Fig. S3). During 2008, hourly gas emissions ranged from 0.39 to 0.53 g h−1 (Fig. S4), with a minimum in July and maximum in January. Apart from the water PCB concentrations, wind speed was one of the most important factors affecting PCB air–water exchange (Achman et al., 1993).

2.4. In situ airborne PCB samples

To examine the influence of IHSC emissions on actual airborne PCB concentrations, we deployed air samples in three locations. During the second week of August 2006, two Hi-Vols were deployed in an intensive field campaign above the waters of IHSC at the harbor and canal to estimate the air–water flux of PCBs from this waterway. Results of these over-water measurements have been previously reported (Martinez et al., 2010). The Hi-Vols ran in parallel during the day, collecting integrated samples over 10–12 h. In 2008 two Hi-Vols were deployed at two locations near IHSC to sample monthly and annual variability in concentrations (Fig. S5). The first Hi-Vol (Hi-Vol1) was located at the Lake George Branch, ~50 m west of the canal, with the purpose of measuring the concentrations generated by the canal. The second Hi-Vol (Hi-Vol2) was located at East Chicago Central High School, ~600 m south of Lake George Branch, to evaluate potential population exposure to the school’s 1500 students. Both samplers ran generally one day per month, from 6:30 am to 6:30 pm. Hi-Vol1 ran from January to May and again in July, and Hi-Vol2 from January to June and from August to December. Because PCBs in air partition ca. 95% to the gas phase (Sun et al., 2006; Wethington and Hornbuckle, 2005), only gas-phase PCBs obtained from the XAD-2 resin was analyzed. Particle-bound PCBs were collected via quartz fiber filters, but not used for this study.

2.5. Analytical method

The complete analytical method for PCB congeners in XAD-2 is described elsewhere (Hu et al., 2008; Martinez et al., 2010). Details of analytical methods for PCBs, as well as our quality control protocol and results, including laboratory blanks, limits of quantification, surrogate recoveries and accurate assessed with standard reference materials, are provided in the Supplementary material. We employed Kruskal–Wallis non-parametric one way analysis variance for model and measured samples, cluster analysis (Principal Component Analysis, PCA) and similarity analysis (cosine theta, cos θ) for PCB congener signals. Details of the statistical methods are described in the Supplementary material.

3. Results and discussion

3.1. AERMOD evaluation. August 2006

Modeling results from August 2006 using the eight meteorological data sets were utilized to assess the performance of AERMOD during the intensive field study and select the most appropriate meteorological data to apply to a full year study for 2008.

Meteorological data yielded significant differences in the modeled 1 h above water ΣPCB concentrations for August 7th–12th 2006. Modeled concentrations ranged from 0 to 16,000 pg m−3, depending on the meteorological data used (Fig. 1). Statistically significant differences in the 1 h modeled concentrations were found between each of the 8 meteorological inputs (Kruskal–Wallis, p < 0.001). However, if a 24 h average was used, no significant difference was found (Kruskal–Wallis, p > 0.05). The average concentration variability between the eight meteorological data yielded 81% (RSD), which is not as high as expected given the major structural and geospatial differences in meteorology from IDEM, WRF, and local observations.

Fig. 1.

Fig. 1

Open in a new tab

Measured and modeled PCB concentration using IDEM, WRF and observed data above the waters of IHSC during August 7–12, 2006. Lines and symbols represent 1 h and 10 h average, respectively. Note differences in y-axis scales.

Not only were the hourly concentrations different; the temporal patterns were also distinct for each approach. The IDEM data from South Bend and Lincoln yielded high spikes during the night, corresponding to the pronounced diel pattern in planetary boundary layer height and wind speed, with nighttime concentrations 5–30 times higher than daytime concentrations. Simulations driven by WRF were influenced by the Lake’s invariant marine boundary layer, with only one episodic peak, while simulations using observed data from Gary airport yielded higher values during the day than at night.

We assessed the meteorological parameters that could affect the modeled concentrations of PCBs in the air. Only wind speed and mixing height from the IDEM data lead to major differences in modeled concentrations (Fig. S6).

Similar to the hourly modeled concentrations, significant differences were found for the 10 h average modeled concentrations (Kruskal–Wallis, p < 0.001) for the same field sampling period, with the exception between IDEM and WRF results (Kruskal–Wallis, p > 0.05). In contrast to the hourly values, all approaches resulted in very similar temporal patterns (Fig. 1).

Contributions from IHSC emissions to the observed above water ΣPCB concentrations ranged in average from 10% to 17%, depending on the meteorological data used. The lowest contribution was obtained when using the Gary observations set, and the highest from WRF. IDEM data yielded a 15% contribution. These findings suggest some combination of three potential conclusions: (i) emissions from IHSC are underestimated; (ii) other local and regional PCB sources dominate observed concentrations above IHSC; or (iii) the available meteorological data and AERMOD are not capable of resolving the highly localized boundary layer dynamics and vertically resolved boundary layer wind profiles at a coastal industrial site, which may not be adequately simulated by either a near-shore modeled grid cell nor an inland site. Considering the known high PCB concentrations found upwind in Chicago (Hu et al., 2010), the many unknown local industrial sources in and around East Chicago, and the high likelihood that other soils, Lake Michigan, and the built environment are also secondary sources to the air, it is most probable that other sources impacted the samples, and that the estimated percentage contributions represent a realistic source contribution.

In addition to the results from August 2006, we evaluated AERMOD using both IDEM and WRF meteorological data sets for 2008 (Fig. 2 and Fig. S7). Above water modeled concentrations yielded very similar patterns and concentrations for the first half of the year. However, during the second half of the year, results using WRF show greatly reduced concentrations, with the exception of one October episode. This result is consistent with the high wind speeds and deep marine boundary layer over the warm lake during the fall and winter (Spak, 2009; Spak and Holloway, 2009). Significant difference was found for 2008 hourly,12 h and 24 h average modeled concentrations (Kruskal–Wallis, p < 0.001), but not for monthly averages (Kruskal–Wallis, p > 0.05). The variability was 58% (RSD), lower than for August 2006.

Fig. 2.

Fig. 2

Open in a new tab

Modeled 1h ΣPCB concentration (pg m−3) for 2008 above the water in the harbor (top panel), Lake George Branch site (middle panel), and East Chicago Central High School (bottom panel). Concentrations were modeled at 1 m above the water/ground. IDEM was used as meteorological data. See the different scale in the y-axis.

Since the obtained modeled concentrations for August 2006 and 2008 do not support any one set of meteorological data, IDEM data was selected to be applied for 2008 because as the current regulatory configuration, and also because monthly (August 2006) as well as annual (2008) results were similar to simulations using WRF, especially for 24 h to seasonal average concentrations.

3.2. 2008 Modeled PCB concentrations

As expected, higher hourly ΣPCB concentrations were simulated above water than at the other two locations. There is no clear annual concentration trend above water, although there are local maxima during the months of February, May, and September. Fig. 2 shows modeled 1 h PCB concentrations above water, Lake George Branch site (Hi-Vol1) and East Chicago Central High School (Hi-Vol2) for 2008 using meteorology provided by IDEM. High hourly modeled concentrations were found, with annual 1 h maxima as high as 46,000 pg m−3. Modeled geometric mean (GSD) hourly concentrations above water, at the Lake George Branch, and East Chicago Central High School yielded 1200 (3.0), 170 (16), and 19 (31) pg m−3, respectively.

Generally, IHSC emissions do not impact East Chicago Central High School and residential areas of East Chicago, but when they do, PCB concentrations generated by IHSC are relative high values (>1000 pg m−3). To estimate the impact of IHSC to the local atmospheric PCB concentration, we evaluated hourly model concentrations at Central High School. This location was chosen because children spend ca. 90 h per year outdoors during school time (recess is ca. 30 min per day, with ca. 180 days per year), and the close proximity suggests exposure to PCBs from IHSC (Ampleman et al., 2015). Results showed that 80% of the time (290 days), modeled concentrations due to emissions from IHSC were 0.0 pg m−3, as the wind rarely blows from the northeast at this location. However, for short periods of time when the wind was blowing from the IHSC to Central High School, the modeled concentrations yielded higher values. For example, 10% of the time (37 days) concentrations in excess of 180 pg m−3 were modeled, 7% of the time (24 days) 1800 pg m−3, and 18,000 pg m−3 was modeled for 17 h (0.2%) (Fig. S7). We note that these air concentrations are far below air quality performance standards developed for particular projects, such as the Hudson River dredging project. For the dredging in the Hudson River, a concern level of 80,000 pg m−3 and a 24 h average standard of 110,000 pg m−3 were implemented by U.S. EPA for residential areas (Parsons, 2009; USEPA, 2004b).

3.3. Measured and modeled PCB concentrations (2008)

ΣPCB concentrations measured near the canal (Hi-Vol1) ranged from 140 to 6500 pg m−3, with a median of 890 pg m−3 (interquartile range, IQR, 430-1300). Concentrations at the high school (Hi-Vol2) ranged from 70 to 1700 pg m−3, with a median of 840 pg m−3 (IQR 110-1400) (Fig. S9). Measured concentrations in both locations showed similar seasonality in concentration until July, with the exception of two high concentrations collected during March at Hi-Vol1. A small increase in the concentrations from January to June was observed, staying constant until November, when it dropped again in December to similar levels detected in January and February (Hi-Vol2). These samples did not show a concentration–temperature dependence, as commonly reported for PCBs in urban areas (Hu et al., 2010; Totten et al., 2006). Three samples from Hi-Vol1 and one from Hi-Vol2 contained an unusually elevated amount of PCB3 (4-monochlorobiphenyl), which is discussed later (Fig. 3). A statistical difference in concentrations was found between the two locations (Kruskal-Wallis, p < 0.05 for samples from the same month), but if the high PCB3 samples were removed, there was no statistical difference in ΣPCB concentrations at the two locations (Kruskal-Wallis, p > 0.05). Individual PCB congener concentrations for each sample are presented in tables S2 and S3.

Fig. 3.

Fig. 3

Open in a new tab

PCB congener profile of modeled average monthly emissions from 2008 (top panel), average samples without enriched PCB3 samples (middle panel) and average enriched PCB3 samples (bottom panel). Not all congeners detected are labeled in the X-axis. Y-axis represents the mass fraction of total PCBs. See the different scale in the y-axis.

AERMOD predictions enabled us to better understand the contribution and impact of IHSC to the surrounding community, and perhaps identify other source(s). Direct comparison between the modeled and observed concentrations at the two locations for the sampling periods revealed that emissions from IHSC, as determined from AERMOD modeling, could not account for the measured concentrations. The predicted concentrations were less than 85% of the observed concentrations, and only on one occasion did AERMOD overestimate the observed concentration, by 160% (Hi-Vol1, 02/01/2008). On average, the contribution from IHSC to the PCBs measured at the two locations was 10%. Table S5 summarizes modeled and observed concentrations at Hi-Vol1 and Hi-Vol2 during the sampling periods.

If we consider wind direction, new evidence of other PCB source(s) larger than IHSC emerges. The winds seldom blew toward our samplers during the sampling periods. Of the 21 samples we collected, only 43% of the samples (9 samples) were impacted by the plume for at least one hour. Of those 9 samples, wind blew from IHSC towards the samplers less than half of the time, on average.

We estimated the contribution of emitted PCBs at the Lake George Branch site on a monthly basis and found that contribution to be 11% if all samples are considered, and 25% if we exclude the four samples that seem to have a distinct PCB3 source. At the high school site, the average monthly contribution from IHSC emission is 9%. We conclude that there are other important sources of airborne PCBs in the surrounding area, such as contaminated soils, industrial sites, building materials, and even indoor air. To our knowledge, no such sources have been reported for the area, despite their likelihood in this city with a long history of heavy industrial activity.

3.4. PCB congener profile analysis

The PCB congener profile of a sample can be a useful tool for investigating relationships between different samples. Airborne PCB congener profiles have been linked with specific commercial Aroclor mixtures as an indicator of the major sources (Asher et al., 2012; Henry and Christensen, 2010; Praipipat et al., 2013; Rodenburg et al., 2011; Totten et al., 2006). We examined the PCB congener profiles of IHSC emission (Fig. 3), the air at the two East Chicago sites (Figs. S11 and S12), IHSC air and water from 2006 (Fig. S13), and five Aroclors. As expected, the emission congener profile is very similar to the water PCB profile (cos θ = 0.99), whereas the average PCB congener profile from the air samples collected above water was slightly different (cos θ = 0.88). A Cos θ value of 0.0 describes two completely different profiles and 1.0 describes two identical profiles (DeCaprio et al., 2005).

The air PCB signature contained more middle and high molecular height congeners, which were missing in the emission and water profiles. The PCB congener signal of the air samples from the two East Chicago sites also strongly resembled the emissions signal, with four exceptions. Several air samples showed a very different profile from the rest. PCA resulted in two clearly grouped clusters. One cluster contained the samples, emission profile and the Aroclors, with the exception of Aroclor 1221. The second cluster included four samples that were found to be enriched in PCB3 (Fig. 4). We note that the enriched PCB3 signal was found in only these four samples collected in late February and March. This unusual PCB congener pattern was not apparent in samples collected in the other months.

Fig. 4.

Fig. 4

Open in a new tab

PCA for all the Hi-Vol samples (white circles), Aroclors commercial mixtures (red circles) and IHSC emission signal (cyan circle). PCB congener profiles were used in this analysis. Cluster A groups all the samples, Aroclors (1016, 1242, 1248 and 1254) and the emission signal, except for Aroclor 1221 and samples with high value of PCB3 (Cluster B). Number in parentheses represents the variance for each principal component. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

We further evaluated the statistical similarity of each PCB signal against every other one using the Cos theta (Table S6) (DeCaprio et al., 2005). This approach evaluates the correlation between two signals and the outcome reinforced our PCA finding that the four samples in group B are very different than those in group A and that most of the air samples strongly resemble the PCB emission signal (Fig. 3). Overall, these findings demonstrates the difficulties in establishing associations between PCB sources and air samplers in a system as East Chicago, and clearly suggest that there are other PCB sources in the area with similar and dissimilar PCB congener profile as IHSC emission.

3.5. PCB 3 (4-monochlorobiphenyl)

Most of the measured samples were enriched in the di-to the penta-chlorinated congeners, with a few interesting exceptions. For example, samples collected at the end of February to the end of March showed a strong PCB3 signal at the Hi-Vol1 location (02/20/ 2008, 03/01/2008, and 03/21/2008). This was also true for the sample collected at the end of February at the Hi-Vol2 location (02/ 26/2008). PCB3 represented more than 35% of the total mass of ΣPCBs in those samples (Fig. 3). PCB3 has been implicated as both a human toxicant and a marker of inhalation exposure to PCBs (Dhakal et al., 2013; Ludewig et al., 2008; Xie et al., 2010).

IHSC is not the source of PCB3. The harbor water and sediments, which were originally contaminated with Aroclor 1248, do not contain significant amounts of this congener. In fact, no Aroclor mixture contains this much PCB3 relative to other congeners. Aroclors 1221 and 1232 are enriched in PCB3, 24% and 14% by weight, respectively (Sax and Isakov, 2003). However, the other high content congeners (PCBs20 + 28, PCB52) detected in the enriched PCB3 samples do not match the profiles of Aroclors 1221 and 1232. It is possible that the signature of the enriched PCB3 samples is due to a combination of Aroclors 1016 (PCBs20 + 28), 1221 (PCB3) and/or 1232 (PCB3), but the high percentage of PCB52 found in the samples (ca. 10%) is not found in any of the Aroclor mixtures. The only known non-Aroclor sources high in PCB3 are cement kilns, which have been shown to emit PCBs to the atmosphere, and PCB3 can be as high as 22% of total emissions (Ishikawa et al., 2007). The nearest cement plant with a cement kiln is more than 100 km from the sampling sites. We therefore suspect that these facilities are not the source of PCB3 that we detected in our samples.

No clear wind pattern was found for the four PCB3 enriched samples (Fig. S14). There is also no correlation with temperature, suggesting that the source is not due to volatilization from surfaces. We conclude that the source is unlikely to be a result of legacy disposal of Aroclors or an industrial process. Thus, we cannot speculate on what that source(s) might be.

3.6. Uncertainties

Sources of uncertainty in this study include PCB emissions, meteorological data and AERMOD’s model structure and configuration (Sax and Isakov, 2003). PCB emissions calculated here used measured environmental factors (wind speed, water temperature, etc.), laboratory or estimated physical–chemical properties, and measured water PCB concentrations. We have previously shown using a Monte Carlo simulation, that even though we applied an error distribution to all the parameters involve in the emission calculations, including an error of 20% in the water concentration, the final PCB emission rates changed by less than 3% (Martinez et al., 2010). This change in emissions resulted in less than a 5% variation in the final modeled airborne PCB concentrations.

Thus, local meteorological data is the most uncertain component in estimating dispersion from passive area sources, particularly for hourly predictions. However, if we consider annual, monthly or even 24 h averages, there is no significant difference in results between the meteorological approaches and data sets used. Analysis of measured and predicted over-water PCB concentrations (August 2006) showed a significant difference: the variability from the 8 meteorological approaches resulted in an average of 81 ± 45%. Hourly 2008 using IDEM and WRF data yielded an average RSD for the same location of 58 ± 40%. These results indicate that our prediction is accurate to about a factor of two.

PCBs in air occur almost completely in the gas phase (ca. 95%) (Wethington and Hornbuckle, 2005). Thus, particulate as well as wet depositions represent small contributions to ΣPCB deposition in relation to gas. Totten et al. reported for the Hudson River Estuary a proportion of gaseous, dry and wet deposition fluxes of ca. 80%, 12% and 8%, respectively (Totten et al., 2004). Wethington and Hornbuckle showed that ca. 90% of the total atmospheric PCB deposition from Milwaukee to Lake Michigan was due to gas-phase deposition (Wethington and Hornbuckle, 2005). Hence, the omission of the particulate and the wet depositions would affect our results by less than 20%.

4. Conclusions

Results from this study support the use of contemporary regulatory dispersion models in estimating local atmospheric SVOC concentrations from contaminated waterways and other secondary area sources. At present, the major challenge of estimating this type of dispersion is in estimating emissions rates. In the case of SVOCs, where long-term average concentrations are the metric of concern for health and ecosystem impacts, the results showed here indicate that even 24 h averages using widely different observed and model meteorological data sets do not yield differences in the modeled concentrations.

Although IHSC contributions to the local PCB concentration in the atmosphere found in this study seems low (10%–15% on an annual basis), their relevance increases more due to the proximity of schools and people living near IHSC. Indeed, we modeled episodes of high concentration that are about twenty times background levels in East Chicago Central High School due only to the emissions from IHSC. This investigation supports the need to adequately remediate IHSC. Currently, IHSC is undergoing a 30-year navigational dredging project, and it is not clear if airborne PCB emissions will increase or decrease once dredging is concluded. The uncertainty is compounded by the fate of the contaminated sediments, which will be held in a CDF next to IHSC.

Supplementary Material

1

HIGHLIGHTS.

  • AERMOD was used to model atmospheric dispersion of PCB emissions from IHSC.

  • Emissions from IHSC contributed ca. 15% of the observed concentrations above water, and ca. 10% at a nearby school.

  • Exposure estimates from annual to 24 h averages independent of data sources for observed and modeled coastal meteorology.

  • Observed enriched PCB3 samples suggest a nearby non-Aroclor source.

Acknowledgments

This work was funded as part of the Iowa Superfund Research Program, NIEHS Grant P42ES013661, and by Grant UL1RR024979 from the National Center for Research Resources, NIH. At the University of Iowa, we thank our laboratory director Collin Just and the undergraduate students Kristen Isley and Sean Nichols for their help in the laboratory, and Dr. Kai Wang for statistics advice. We also thank Salvador Ramirez for collecting the air samples in East Chicago.

Appendix A. Supplementary data

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.atmosenv.2015.10.040.

Footnotes

The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIEHS or the NIH.

References

  1. Achman DR, Hornbuckle KC, Eisenreich SJ. Volatilization of poly-chlorinated biphenyls from Green Bay, Lake Michigan. Environ Sci Technol. 1993;27:75–87. [Google Scholar]
  2. Ampleman MD, Martinez A, DeWall J, Rawn DFK, Hornbuckle KC, Thorne PS. Inhalation and dietary exposure to PCBs in urban and rural cohorts via congener-specific measurements. Environ Sci Technol. 2015;49:1156–1164. doi: 10.1021/es5048039. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Asher BJ, Ross MS, Wong CS. Tracking chiral polychlorinated biphenyl sources near a hazardous waste incinerator: fresh emissions or weathered revolatilization? Environ Toxicol Chem. 2012;31:1453–1460. doi: 10.1002/etc.1852. [DOI] [PubMed] [Google Scholar]
  4. Basu I, Hafner W, Mills W, Hites R. Differences in atmospheric persistent organic pollutant concentrations at two locations in Chicago. J Gt Lakes Res. 30:310–315. [Google Scholar]
  5. Chiarenzelli JR, Scrudato RJ, Wunderlich ML. Volatile loss of PCB Aroclors from subaqueous sand. Environ Sci Technol. 1997;31:597–602. [Google Scholar]
  6. Cimorelli AJ, Perry SG, Venkatram A, Weil JC, Paine RJ, Wilson RB, Lee RF, Peters WD, Brode RW. AERMOD: a dispersion model for industrial source applications. Part I: general model formulation and boundary layer characterization. J Appl Meteorol. 2005;44:682–693. [Google Scholar]
  7. Connolly JP, Zahakos HA, Benaman J, Ziegler CK, Rhea JR, Russell K. A model of PCB fate in the upper Hudson River. Environ Sci Technol. 2000;34:4076–4087. [Google Scholar]
  8. Davis N, Arumachalam S, Brode R. MCIP2AERMOD: a prototype tool for preparing meteorological inputs for Aermod. 7th Annual Models-3 CMAS Users Conference CMAS.2008. [Google Scholar]
  9. DeCaprio AP, Johnson GW, Tarbell AM, Carpenter DO, Chiarenzelli JR, Morse GS, Santiago-Rivera AL, Schymura MJ. Polychlorinated biphenyl (PCB) exposure assessment by multivariate statistical analysis of serum congener profiles in an adult Native American population. Environ Res. 2005;98:284–302. doi: 10.1016/j.envres.2004.09.004. [DOI] [PubMed] [Google Scholar]
  10. Dhakal K, Adamcakova-Dodd A, Lehmler HJ, Thorne PS, Robertson LW. Sulfate conjugates are urinary markers of inhalation exposure to 4-Chlorobiphenyl (PCB3) Chem Res Toxicol. 2013;26:853–855. doi: 10.1021/tx4001539. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Fry J, Xian G, Jin S, Dewitz J, Homer C, Yang L, Barnes C, Herold N, Wickham J. Completion of the 2006 National Land Cover Database for the conterminous United States. PE&RS. 2011;77(9):858–864. [Google Scholar]
  12. Gesch D, Oimoen M, Greenlee S, Nelson C, Steuck M, Tyler D. The national elevation dataset. Photogramm Eng Remote Sens. 2002;68 [Google Scholar]
  13. Harner T, Mackay D, Jones KC. Model of the long-term exchange of PCBs between soil and the atmosphere in the south U.K. Environ Sci Technol. 1995;29:1200–1209. doi: 10.1021/es00005a010. [DOI] [PubMed] [Google Scholar]
  14. Henry RC, Christensen ER. Selecting an appropriate multivariate source apportionment model result. Environ Sci Technol. 2010;44:2474–2481. doi: 10.1021/es9018095. [DOI] [PubMed] [Google Scholar]
  15. Hillery BR, Basu I, Sweet CW, Hites RA. Temporal and spatial trends in a long-term study of gas-phase PCB concentrations near the Great Lakes. Environ Sci Technol. 1997;31:1811–1816. [Google Scholar]
  16. Hu D, Hornbuckle KC. Inadvertent polychlorinated biphenyls in commercial paint pigments. Environ Sci Technol. 2010;44:2822–2827. doi: 10.1021/es902413k. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Hu D, Lehmler HJ, Martinez A, Wang K, Hornbuckle KC. Atmospheric PCB congeners across Chicago. Atmos Environ. 2010;44:1550–1557. doi: 10.1016/j.atmosenv.2010.01.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Hu D, Martinez A, Hornbuckle KC. Discovery of non-aroclor PCB (3,3′-Dichlorobiphenyl) in Chicago air. Environ Sci Technol. 2008;42:7873–7877. doi: 10.1021/es801823r. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Indiana Department of Environmental Management. [accessed 10.10.14];Air Dispersion Meteorological Data. http://www.in.gov/idem/airquality/2376.htm.
  20. International Joint Commission. Final. 2003. Status of Restoration Activities in Great Lakes Areas of Concern: a Special Report. [Google Scholar]
  21. Ishikawa Y, Noma Y, Mori Y, Sakai S. Congener profiles of PCB and a proposed new set of indicator congeners. Chemosphere. 2007;67:1838–1851. doi: 10.1016/j.chemosphere.2006.05.075. [DOI] [PubMed] [Google Scholar]
  22. Ludewig G, Lehmann L, Esch H, Robertson LW. Metabolic activation of PCBs to carcinogens in vivo – a review. Environ Toxicol Pharmacol. 2008;25:241–246. doi: 10.1016/j.etap.2007.10.029. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Martinez A, Wang K, Hornbuckle KC. Fate of PCB congeners in an industrial harbor of Lake Michigan. Environ Sci Technol. 2010;44:2803–2808. doi: 10.1021/es902911a. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Moore G, Paine R, Hienold D. Limitations of steady-state dispersion models and possible advanced approaches. Air & Waste Management Association 105th Annual Meeting and Exhibition; San Antonio, TX. 2012. [Google Scholar]
  25. Morselli M, Ghirardello D, Semplice M, Raspa G, Di Guardo A. Integration of an atmospheric dispersion model with a dynamic multimedia fate model: development and illustration. Environ Pollut. 2012;164:182–187. doi: 10.1016/j.envpol.2012.01.039. [DOI] [PubMed] [Google Scholar]
  26. National Center for Atmospheric Research. User’s Guide for the Advanced Research WRF (ARW) Modeling System Version 3.3. 2012. [Google Scholar]
  27. Parsons. Hudson River PCBs Superfund Site. GE Company; 2009. Phase 1 Remedial Action Community Health and Safety Plan. [Google Scholar]
  28. Perry SG, Cimorelli AJ, Paine RJ, Brode RW, Weil JC, Venkatram A, Wilson RB, Lee RF, Peters WD. AERMOD: a dispersion model for industrial source applications. Part II: model performance against 17 field study databases. J Appl Meteorol. 2005;44:694–708. [Google Scholar]
  29. Petrich NT, Spak SN, Carmichael GR, Hu D, Martinez A, Hornbuckle KC. Simulating and explaining passive air sampling rates for semivolatile compounds on polyurethane foam passive samplers. Environ Sci Technol. 2013;47:8591–8598. doi: 10.1021/es401532q. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Praipipat P, Rodenburg LA, Cavallo GJ. Source apportionment of poly-chlorinated biphenyls in the sediments of the Delaware River. Environ Sci Technol. 2013;47:4277–4283. doi: 10.1021/es400375e. [DOI] [PubMed] [Google Scholar]
  31. Rodenburg LA, Du S, Xiao B, Fennell DE. Source apportionment of polychlorinated biphenyls in the New York/New Jersey Harbor. Chemosphere. 2011;83:792–798. doi: 10.1016/j.chemosphere.2011.02.058. [DOI] [PubMed] [Google Scholar]
  32. Sax T, Isakov V. A case study for assessing uncertainty in local-scale regulatory air quality modeling applications. Atmos Environ. 2003;37:3481–3489. [Google Scholar]
  33. Sinkkonen S, Paasivirta J. Degradation half-life times of PCDDs, PCDFs and PCBs for environmental fate modeling. Chemosphere. 2000;40:943–949. doi: 10.1016/s0045-6535(99)00337-9. [DOI] [PubMed] [Google Scholar]
  34. Spak S. Characterizing Chemical Transport of Ozone and Fine Particles in the Great Lakes Region. University of Wisconsin-Madison; 2009. [Google Scholar]
  35. Spak SN, Holloway T. Seasonality of speciated aerosol transport over the Great Lakes region. J Geophys Res. 2009;114:D08302. [Google Scholar]
  36. Sun P, Basu I, Hites RA. Temporal trends of polychlorinated biphenyls in precipitation and air at Chicago. Environ Sci Technol. 2006;40:1178–1183. doi: 10.1021/es051725b. [DOI] [PubMed] [Google Scholar]
  37. Totten LA, Gigliotti CL, Offenberg JH, Baker JE, Eisenreich SJ. Reevaluation of air-water exchange fluxes of PCBs in Green Bay and southern Lake Michigan. Environ Sci Technol. 2003;37:1739–1743. doi: 10.1021/es026093x. [DOI] [PubMed] [Google Scholar]
  38. Totten LA, Gigliotti CL, Vanry DA, Offenberg JH, Nelson ED, Dachs J, Reinfelder JR, Eisenreich SJ. Atmospheric concentrations and deposition of polychorinated biphenyls to the Hudson River Estuary. Environ Sci Technol. 2004;38:2568–2573. doi: 10.1021/es034878c. [DOI] [PubMed] [Google Scholar]
  39. Totten LA, Stenchikov G, Gigliotti CL, Lahoti N, Eisenreich SJ. Measurement and modeling of urban atmospheric PCB concentrations on a small (8 km) spatial scale. Atmos Environ. 2006;40:7940–7952. [Google Scholar]
  40. USEPA. Users Guide for the AERMOD Terrain Preprocessor (AERMAP) 2004a. [Google Scholar]
  41. USEPA. Hudson River PCBs Superfund Site Quality of Life Performance Standards. 2004b. [Google Scholar]
  42. USEPA. AERSURFACE User’s Guide. 2008. [Google Scholar]
  43. USEPA. AERMOD Modeling System. 2014. [Google Scholar]
  44. Wethington DM, Hornbuckle KC. Milwaukee, WI, as a source of atmospheric PCBs to Lake Michigan. Environ Sci Technol. 2005;39:57–63. doi: 10.1021/es048902d. [DOI] [PubMed] [Google Scholar]
  45. World Health Organization. Indoor Air Quality: Organic Pollutants. EURO Reports and Studies 111; Report on a WHO Meeting; Berlin. 23–27 August 1987; Copenhagen: World Health Organization Regional Office for Europe; 1989. [Google Scholar]
  46. Xie W, Wang K, Robertson LW, Ludewig G. Investigation of mechanism(s) of DNA damage induced by 4-monochlorobiphenyl (PCB3) metabolites. Environ Int. 2010;36:950–961. doi: 10.1016/j.envint.2009.12.004. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1
NIHMS735453-supplement-1.docx (4.7MB, docx)

RESOURCES