Spatial Distribution, Air-Water Exchange, and Toxicity of Organic Pollutants Using Passive Samplers

Thousands of hydrophobic organic contaminants (HOCs) are present in air and water worldwide, yet we know little about how these chemicals’ concentrations vary spatially and temporally, or what biological effects they have in concert. The first four studies described in this dissertation present data from a sampling campaign in which passive polyethylene samplers (PEs) deployed throughout the lower Great Lakes region (Lake Erie and Lake Ontario) from 2011 to 2014. Results were used to deduce air-water fluxes and analyze spatial trends of the truly gaseous and dissolved fraction of three distinct groups of HOCs: polycyclic aromatic hydrocarbons (PAHs), polycyclic musks (PCMs), and organic flame retardants (OFRs), with the goal of better understanding how sources and physico-chemical properties determine the environmental transport and spatial distribution of these HOCs. The specific objectives of these studies were to determine whether gaseous and dissolved HOCs exhibited positive correlation with regional population density within 25 km of each site in the lower Great Lakes region, investigate whether diffusive air-water exchange of HOCs was primarily leading to volatilization from, or absorption into, the lakes’ surface waters, and investigate health risks of ambient urban air by measuring aromatic hydrocarbon receptor (AhR)-mediated potency of the truly gaseous mixture of HOCs accumulated in PEs deployed in air on the Lake Erie shoreline near Cleveland (OH). Results showed that the radius at which strongest correlation between gaseous HOC concentration and human population was observed depended on vapor pressure, and a relationship between the maximum distance where significant correlation occurred and compound vapor pressure is presented for amassed PAH, PBDE, and PCM data. Air-water exchange calculations based on simultaneously deployed air and water PEs indicated that diffusive exchange of PAHs was variable based on compound and season. PCMs were found to be volatilizing from the lakes’ surface waters, suggesting that Lake Erie and Lake Ontario were acting as secondary sources of PCMs, while PBDEs were absorbed into surface waters. Bioassay experiments performed on PE extracts showed that <30% of AhR-mediated potency for gaseous air extracts was explained by target compounds measured via chemical analysis, suggesting that targeted analysis may underestimate health risks posed by gas-phase ambient air. The fifth and sixth studies described in this dissertation focused on measuring uptake of emerging and legacy HOCs into PEs to inform future calculation of ambient air and water concentrations from PE measurements. PE uptake profiles over 21-day deployments were used to determine whether target compounds reached equilibrium during deployment, and PE-water and PE-air partitioning coefficients (KPEW and KPEA) were calculated. KPEW values for PAHs agreed fairly well with empirical values from literature in most cases, while values for PCMs and OPEs were generally lower than predicted based on chemical properties, suggesting that PE-derived concentrations for these compounds may be underestimated when using this approach. The seventh and final study included in this dissertation presents concentrations of dissolved organophosphate esters (OPEs), a group of emerging OFRs with atypical physico-chemical properties, derived from PEs deployed in the North Atlantic deep ocean from 2014-2015 and in Canadian Arctic surface waters during the summers of 2015 and 2016 to investigate long-range transport of OPEs to remote aquatic environments. For the first time, estimated concentrations of OPEs in polar ocean surface water and remote ocean deep water are reported. The greatest concentrations of OPEs were measured in Canadian Arctic surface waters, with the chlorinated OPE species most abundant. OPEs exhibited unexpectedly flat vertical profiles in the North Atlantic Fram Strait, possibly due to a high degree of mixing and/or release of dissolved-phase OPEs from sinking particles. This study demonstrated that OPEs are widespread, even in remote environments, and that concentrations are much greater than those of other OFRs in the Arctic, suggesting that OPEs should be a priority for further study.


LIST OF TABLES
HOCs are of particular concern because they are often persistent, capable of longrange transport, and bioaccumulative. In this dissertation, three types of HOCs representing distinct use patterns and sources were investigated: polycyclic aromatic hydrocarbons (PAHs), polycyclic musks (PCMs), and organic flame retardants (OFRs). lower concentrations in residential and rural areas. [5][6][7] One of the few studies to measure aqueous PAHs in Great Lakes surface waters observed that concentrations were greatest in Lakes Erie and Ontario, where they reached about 5 ng/L. 8 PCMs are synthetic fragrance compounds widely used as additives in personal care products and household cleaners. 9 They are ubiquitous in aquatic environments, with concentrations generally in the 1-1000 ng/L range in rivers and lakes. 10

INTRODUCTION
Polycyclic aromatic hydrocarbons (PAHs) are ubiquitous pollutants that originate from oil spills as well as anthropogenic and natural combustion processes.
Major sources include fossil fuel combustion, metal production, waste incineration, residential and commercial biomass burning, and vehicular emissions. [1][2][3][4][5] PAHs are often associated with densely populated areas, especially in industrialized countries. 3,4,6,7 PAHs and their transformation products are a primary carcinogenic component of urban air pollution and health effects resulting from chronic exposure are a serious concern. 8,9 Polyethylene passive samplers (PEs) are cost-effective, simple tools with lower detection limits than traditional active sampling techniques. Instead of pumping air or water through a filter, PEs accumulate hydrophobic organic contaminants (HOCs) over time via diffusion, accumulating only truly dissolved or gas-phase molecules. 10 Concentrations of truly dissolved HOCs are of interest because this fraction is available for direct diffusive exchange between water and other reservoirs such as air, biota, or sediment.
The use of PEs facilitates simultaneous spatially resolved measurements and calculations of air-water diffusive exchange rates. For most HOCs, concentrations measured by PEs reflect a time-integrated concentration representative of the entire deployment period. For compounds that equilibrate during deployment, concentrations reflect the most recent concentration the sampler was exposed to. PEs have previously been used to measure HOCs in water and air [11][12][13] and to calculate airwater gradients of HOCs, but this method has not been applied to the lower Great Lakes. [14][15][16] Lake Erie and Lake Ontario are the smallest of the Great Lakes by volume and have estimated residence times of 2.7 and 7.5 years, respectively. 17 About 80% of Lake Erie's water is supplied by the Detroit River, which is fed by Lake Huron via Lake St. Claire. Among the Great Lakes, Lake Erie is the shallowest (average depth 19 m), warmest, and most biologically productive. 18,19 Lake Ontario is much deeper (average depth 86 m) and primarily receives water from Lake Erie via the Niagara River. 19 Currents in the Great Lakes are weak (a few cm/s) with complex temporal variability that depends on recent atmospheric conditions. In the summertime, circulation is generally counterclockwise ( Figure S2-3). 20 The lakes are stratified from May through October and well-mixed for the remainder of the year. 20 Heavy urbanization and valuable ecosystems often coincide along the shores of the lower Great Lakes. Atmospheric deposition from urban sources has been identified as a major source of gaseous and particle-bound HOCs to the region's aquatic environment. 2,7,21,22 Concentrations of total atmospheric PAHs have been shown to correlate strongly with population in this region and urban centers have been linked to significantly increased loadings of contaminants to the lakes. 23,24 In some cases, however, the lakes have been found to act as a source of HOCs via revolatilization. 25 Four sites formed an east-west transect along Lake Ontario's southern shore.
The westernmost site, Grimsby (ON), was an offshore buoy monitored by Environment Canada. On Lake Erie, samplers were deployed at nine US shoreline sites and six offshore sites monitored by Environment Canada. Samplers were deployed at the offshore sites once, during late summer. Samplers at Gibraltar Island (OH) and Toledo (OH) were deployed once during late spring/early summer.

Meteorological Information & Site Characteristics. Monthly wind speed averages
during the sampling campaign ranged from 3.8 m/s in July to 6.1 m/s in November, with the greatest average wind speeds offshore of Toledo. Average air temperatures ranged from 7.7 °C in April to 24.3 °C in July and the mean deployment temperature for all sampling periods was 18.6±1.8 °C. Surface water temperatures were generally very similar to air temperatures and ranged from 3.7˚ C (Lake Ontario in May) to 20 25.1˚ C (Lake Erie in July). 28 There were westerly prevailing winds during the sampling campaign for most of the study region ( Figure S2-2, Table S2-3). 29,30 Precipitation and river discharge were lowest during June and July while flows in late spring and early fall were similar. 31 Locations near major rivers are listed in Table S4.
Sample Analysis. All PEs were spiked with deuterated PAHs and extracted twice, each time for 18 hours. Air PEs were extracted with ethyl acetate followed by hexane.
Aqueous PEs were extracted with dichloromethane followed by hexane. Extracts were concentrated to approximately 100 µL and p-terphenyl-d 14 was added as an injection standard. Extracts were analyzed using an Agilent 6890 GC coupled to an Agilent 5973 MSD in electron ionization (EI) selected ion monitoring (SIM) mode.
PAH analysis and quality control procedures are further outlined by Khairy et al. 11 PAH concentrations were corrected for internal standard recoveries (Table S5) and blank-subtracted using the field blank relevant to the sampling site. If no field blank for the site was available, the average concentration from all available field blanks was used. More information on quality assurance and quality control is in the Supporting Information.

Determination of Sampling Rate and Ambient Concentration. The uptake of
HOCs by PEs is described in detail by Lohmann 32 and PE-air partitioning is detailed by Khairy et al. 11 To determine ambient PAH concentrations from concentrations in polyethylene, site-specific sampling rates were estimated via a method adapted from Booij et al. 33 The average air sampling rate was 28±17 m 3 /day and the average aqueous sampling rate was 112±57 L/day. For more details, see the Supporting Information and Tables S1 and S2.
Physico-chemical Parameters. Sampler-matrix partition coefficients used to calculate ambient concentrations for each PAH are listed in Table S7, along with other physico-chemical properties. Temperature-adjusted partition coefficients were obtained using mean temperature during the deployment period for the nearest meteorological buoy or weather station and the modified van't Hoff equation, as in Khairy et al. 11 The enthalpy of vaporization (ΔH vap ) was used to account for K PEAtemperature sensitivity and internal energy of dissolution (ΔU w ) for K PEW -temperature sensitivity.
Population Analysis. Population data for each sampling site are presented in Table   S8. Total population within a circular area with a 1-cell (about 1 km) radius was calculated using the Focal Statistics tool in ArcMap. The process was repeated for larger radii to create a dataset of the total population within 1, 2, 3, 5, 10, 15, 20, 25, Ratios of gaseous Flra/Flra+Pyr were ≥0. 6 and Phn/ΣMPhns>1 at all sites, suggesting that gaseous PAHs were primarily combustion-derived. 34 The two sites in Cleveland consistently displayed the greatest concentrations of gaseous PAHs except retene throughout the deployment season. Retene is often considered to be indicative of wood smoke or pulp/paper mill effluent, as opposed to fossil fuel combustion. 35,36 Retene was greatest west of Cleveland in Sheffield Lake, but even here accounted for less than 0.7% of total gaseous PAHs. In contrast, Ruge found retene to be a significant component of gaseous PAH profiles at many sites on Lake Superior. 37 Principal component analysis (PCA) using the FactoMineR package 38  PAH concentrations in this study were comparable to those measured by Ruge at urban locations along the shore of Lake Superior. 37  Cleveland and Rochester concentrations in this study. 6 Concentrations in this study were lower than those reported for Alexandria, Madrid, or Lake Chaohu, China and greater than concentrations on the Taiwan coast. 11,[41][42][43] Total (aerosol and dissolved) 2-3-ring PAHs near Lake Victoria, East Africa were lower than 2-3-ring gaseous PAHs in Cleveland, but greater than the remainder of the deployment sites. 44 Gaseous PAHs and Population. Sampling sites were classified as urban, semi-urban, rural, or remote based on population within 3 km (Table S8). Mean Σ 15 PAH for each type of site are summarized in Table 1. For both lakes, the greatest concentrations of gaseous PAHs were observed at urban sites. However, Σ 15 PAH was not significantly different based on site classification using a one-way analysis of variance (ANOVA; p > 0.05). There were no obvious changes in PAH profile composition based on whether the site was urban, semi-urban, rural, or remote ( Figure   S2-4A).
To explore relationships with population in more detail, population within discrete radii of 1 to 40 km from each site were compared to average atmospheric observed for all measured PAHs at some radius, with retene exhibiting the weakest correlation (r 2 1 km = 0.30 at a radius of 1 km, p = 0.02, SE=0.02). This is most likely due to retene's association with wood smoke, as opposed to fossil fuel combustion. 35,36 Strong correlations suggest that urban centers are a primary source of gaseous PAHs (except retene) in the lower Great Lakes region.
For each PAH, the strength of the correlation between population and concentration varied as we changed the radius used to characterize population at the site ( Figure 2-3). All compounds except retene displayed a bimodal relationship, with two radii of maximum correlation. This relationship was less pronounced for the low molecular weight (LMW) PAHs than HMW PAHs. Strong similarities between correlation profiles (e.g., the 5-6-ring PAHs) suggest similar sources and affinities for transport.
Hafner and Hites suggested that the significance of local sources in determining Great Lakes HOC concentrations varies based on a compound's atmospheric lifetime. 7 The atmospheric lifetimes of gaseous PAHs are determined primarily by susceptibility to hydroxyl degradation and gas-particle partitioning. 7 Anthracene exhibited a distinctly shaped correlation curve with two maxima at radii 25 km (r 2 25 km = 0.77) and 5 km (r 2 5 km = 0.77) (Figure 2-3). Anthracene has a short lifetime (1.5 hrs) with respect to hydroxyl radical degradation relative to other PAHs, which may explain why stronger correlation is observed at short distances than for other 3-ring PAHs, 45,46 but this does not explain the comparable correlation at 25 km.
Acenaphthylene is expected to have a similar lifetime to anthracene (1.6 hrs) 45 and exhibited stronger correlations with more local population than fluorene.
Fluorene is often observed to be more stable with respect to photochemical oxidation than similarly-sized PAHs (average lifetime 22 -26 hrs) 45,47 but more distant sources did not become more significant for this compound due to its longer lifetime.
Fluorene correlated less strongly with population than acenaphthylene at all radii, but the divergence was largest at shorter distances.
Gaseous HMW PAHs are expected to have short atmospheric residence times due to reaction with hydroxyl radicals, which may contribute to the increased relevance of local versus long-range sources that was observed for these compounds. 7 These results suggest that reaction with hydroxyl radicals limited the importance of sources distant from sampling sites.  Figure S2-6. Excluding anthracene, PAHs with p L > 10 -4 Pa were most highly correlated with population within a 20 km radius, while PAHs with p L < 10 -4 Pa were most highly correlated with population within 3 km. Other studies have observed similar values for log(p L ) at which PAHs transition from being primarily gaseous to particle-bound. 26,52,53 While S2-6 highlights maximum correlation, many PAHs exhibited significant correlation with population at both 20 km and 3 km. As shown in Figure 2 ng/L), perhaps due to the weakening of summertime stratification. The site was also within 5 km of a major (~150 million L/day) WWTP ( Figure S2-3A).
Air-Water Exchange. Mass transfer coefficients and flux gradients are listed in Tables S11 and S12 and flux gradients for select PAHs are presented in Figure S2-9.
Mass transfer velocity ranged from 0.   Relative significance of population within 20 km and 3 km. The ratio of r 2 20 km to r 2 3 km correlated well with sub-cooled liquid vapor pressure at mean deployment temperature, suggesting that distant sources were more significant for volatile PAHs.  Information. RSDs between sampling rates calculated for replicate atmospheric samplers ranged from 1% -102% with an average of 32%. Passive sampling rates are known to be affected by meteorological factors like wind speed and temperature 10,11 but no significant relationship was observed between air sampling rates and nearby wind speeds, most likely because sampling rates depended on features of the specific site and regional wind speed values did not capture this local variability. RSDs between rates calculated for replicate aqueous samplers ranged from 10%-36% with an average of 21%.

TABLES AND FIGURES
Few studies were available for direct comparison of R s values. Allan et al.
(2013) estimated riverine sampling rates for 300 cm 2 low-density polyethylene (LDPE) samplers to be 6.4-18.5 L/d. 12 Air sampling rates determined by Ruge (2013) for PEs in Lake Superior air (0.6 -70 m 3 /day) were of similar magnitude and variability to those reported here, though aqueous sampling rates were lower in that study (2 -25 L/day). 13 Air-Water Exchange Rates. The direction of flux was determined from the flux ratio as shown in Equation (2), with values > 0 indicating volatilization and values < 0 indicating absorption into surface waters. ( If a compound was below the detection limit in both air and water, no flux was calculated. The standard deviation of the flux ratio was calculated via error propagation based on sampling rates and analytical repeatability (both assumed to contribute 10% uncertainty), and temperature-corrected partitioning coefficients (assumed 50% and 30% uncertainty for ΔU w and ΔH vap , respectively, as in ). 14 The standard deviation of the flux ratio was used to determine whether 52 flux ratios were significantly different from equilibrium. Ratios that were not significantly different from equilibrium are flagged in data tables.
Total flux of PAHs (ng/m 2 /day) was calculated from the air-water flux gradient and mass transfer coefficient as in Equation (3) Table S8.      Flux ratios calculated from data where one concentration was < DL are marked in red. Flux ratios that were not significantly different from equilibrium after error propagation are grayed out. If both air and water concentrations were < DL, no flux ratio was calculated (NA). Table S2- Flux ratios calculated from data where one concentration was < DL are marked in red. Flux ratios that were not significantly different from equilibrium after error propagation are grayed out. If both air and water concentrations were < DL, no flux ratio was calculated (NA).  Figure S3-1).
Shoreline PEs were deployed by trained volunteers as previously described. 19 Briefly, volunteers hung air PEs inside protective metal bowls at a height of about 1.5 m, and tethered water PEs to an anchored line so that they would be secured about 1 m beneath the water's surface. Offshore and nearshore deployments were carried out by workers at Environment Canada and the Ontario Ministry of the Environment, as described previously by Liu et al. 17 Air PEs were secured in a protective chamber 2 m above the water's surface on a buoy and water PEs were enclosed within a perforated metal cage and secured to the buoy about 4 m below the water's surface. After the PEs were recovered, they were shipped back to the laboratory overnight on ice and frozen until extraction.
Extraction and Analysis. PEs from 56 atmospheric deployments (including 9 overwinter deployments) and 39 aqueous deployments were extracted and analyzed.
All PEs were spiked with labeled PAHs (acenaphthene-d 10 , phenanthrene-d 10 , Percent detection for target compounds is presented in Table S7. HHCB and AHTN were found in 15% and 68% of all shoreline air PEs deployed in this study and in 38% and 54% of offshore/nearshore air PEs. In water, HHCB and AHTN were found in 45% and 60% of shoreline PEs and in 47% and 79% of offshore PEs.

Physico-Chemical Properties. Physico-chemical properties of all target analytes and
PRCs are presented in Table S8. PE-air partitioning coefficients (K PEA ) were determined from regression with subcooled liquid vapor pressure as in Khairy and Lohmann. 15 PE-water partitioning coefficients (K PEW ) were calculated from solubility as in Lohmann. 21 K PEA , K PEW , and diffusivity in air (D a ) and water (D w ) for each compound were corrected for each deployment's mean temperature, as detailed further in the SI.  Table S1), l PE is half the PE thickness, K PEM is the PE-matrix partitioning coefficient, and k o is the mass transfer coefficient, which represents the reciprocal sum of PE-side resistance (k PE -1 ), which is dependent on D PE and l PE , and environmental matrix-side resistance (k m -1 ), which is dependent on D a or D w and δ DBL . Best-fit δ DBL values were used to estimate f reached by each PCM during each deployment.

Sampling Rates and Ambient
(1) 90 Average f values for each PCM are presented in Table S9 and show that HHCB and AHTN generally reached > 95% equilibrium in both air and water.
Average δ DBL s for air boundary layers (δ ABL ) were lower for offshore/nearshore PEs the ordinary least-squares linear modeling function (lm) in R. 24 Linear models were 91 further refined using the robust linear model (rlm) function in the MASS package in R, 25 which iteratively fits data to a linear model, weighting outliers depending on their distance from the best-fit line. All presented relationships were found to be statistically significant (p < 0.01) using both approaches. Results were plotted using R package ggplot2. 26

Air-Water Exchange Calculations. Thirty-two pairs of codeployed air and water
PEs were used to investigate time-integrated air-water exchange fluxes. The direction of exchange was determined by calculating the ratio of fugacity in water to fugacity in air (f w /f a ) as in Equation 3, where C ∞,w and C ∞,a represent the concentration of the compound in the PE once it has reached equilibrium with surrounding water and air, respectively. ( A value of f w /f a > 1 indicates volatilization, while f w /f a < 1 indicates absorption. In cases where the concentration in both air and water were <DL, no fugacity ratio was calculated. In cases where the concentration in one medium was <DL, but was >DL in the other medium, a fugacity ratio was calculated by replacing the <DL value with the DL value, as this resulted in the most conservative estimate for the fugacity ratio (see Figure S3-2).
Values for C ∞,w and C ∞,a were determined by correcting the concentration in the PE (C PE ) using the calculated percent equilibrium (f) reached by each compound during deployment. In most cases for AHTN and HHCB, C ∞ ~ C PE because they equilibrated during deployment. The uncertainty in the fugacity ratio was calculated by propagating the uncertainty in the parameters used to calculate C ∞,a and C ∞,w , which 92 is detailed further in the SI. In cases where the fugacity ratio was within one standard deviation from equilibrium, it was not considered significantly different from equilibrium and no flux was calculated.
Air-water exchange fluxes (F a/w ) were calculated using an approach based on the Whitman two-film model 27 as described in Schwarzenbach et al. 28 with wind speed's effect on water-side mass transfer determined using a Weibull distribution to account for the nonlinearity of the effect of wind speed on mass transfer. 29 The mass was calculated using the uncertainty of the parameters used to calculate C ∞,a , C ∞,w , and K PEW,T2 , and assuming 30% relative uncertainty in v a/w . 31 Calculations and error propagation are detailed further in the SI. Average dissolved PCMs are summarized in Table 3-1. Along the southeastern shore of Lake Erie and the northeastern shore of Lake Ontario, concentrations were similar to offshore levels (Ʃ 5 PCM < 100 pg/L) and HHCB was generally <DL, while concentrations were elevated nearer to the urban centers of Toronto and Cleveland and along the southern shore of Lake Ontario. Variation in dissolved Ʃ 5 PCM over multiple deployments is shown in Figure S3 Gaseous PCM Concentrations. Average summertime Ʃ 5 PCM ranged from <DL at sites in Erie (ERI) and Sheffield Lake (SHF) on the southern Lake Erie shoreline, Prince Edward Point (PEP) in northern nearshore Lake Ontario, and eastern offshore Lake Erie (EERI), to 3.2 ng/m 3 in Toledo (TOL). Concentrations of all gaseous PCMs are summarized in Table 3 Table S3-10 and depicted in Figure S3-2. At all sites where HHCB was detected in air and/or water, fugacity ratios suggested it was volatilizing out of surface waters. Fugacity ratios for AHTN also suggested volatilization from surface waters near Toronto and along the southern shore of Lake Ontario, though AHTN was near equilibrium or absorbed into surface waters at some other sites.
The greatest fugacity ratios for both AHTN (f w /f a = 7) and HHCB (f w /f a = 18) were calculated for the PE pair from the late-summer deployment near the mouth of the Oswego River (OSW), during which greater dissolved PCMs were measured than during any other deployment (Ʃ 5 PCM = 4.8 ng/L). Fugacity ratios were generally not significantly different from equilibrium at sites on the southeastern shore of Lake Erie (ERI, DUN, BUF), the northeastern Lake Ontario shoreline/nearshore (CV, PEP, CHB), or at the offshore sites (CERI, EERI).

PE-Derived Air-Water Exchange Fluxes at Non-Steady-State Conditions.
Values of v a/w calculated for HHCB and AHTN ranged from 4.5-8.8 cm/day, which was somewhat slower than rates for PCBs calculated by Liu  in the codeployed PEs at that time. F aw was then compared to F aw,PE by calculating the RPD between the two values. An example from Scenario 2, in which F aw decreased throughout the simulated deployment, is displayed in Figure 3-3. F aw,PE is shown to steadily decline over the deployment along with F aw , but F aw,PE does not capture rapid day-to-day changes in the flux and appears to lag behind F aw by about 20 days. A similar figure is shown for Scenario 1 in Figure S3-5.

99
Each scenario was run 100 times, and each time the RPD between F aw,PE and F aw after 100 days of deployment was recorded. Results are presented in Table 3-3   Volatilization fluxes in this study were driven by elevated dissolved concentrations at shoreline and nearshore sites. These elevated concentrations were expected to be entrained in the nearshore coastal boundary zone, which extends from the shoreline to where the depth of the lake exceeds that of the thermocline. 42 To estimate total losses of dissolved PCMs from the lakes via volatilization, fluxes were averaged over the estimated surface area of the urbanized coastal boundary zone.
The surface area of the Lake Ontario coastal boundary zone was estimated to be 6500 km 2 by extracting the area with depth shallower than 50 m using GIS data from the Great Lakes Commission's Great Lakes Information Network (GLIN), as shown in Figure S3-6. The coastal boundary zone in Lake Erie was more difficult to define, as most of the lake is quite shallow and it does not develop a pronounced seasonal thermocline as does Lake Ontario. From GLIN data, the surface area of Lake Erie shallower than 20 m was estimated to be 15200 km 2 .
Averaging fluxes at all Lake Ontario sites yielded a mean Ʃ 5 PCM flux of 58 ng/m 2 /day over the coastal boundary zone. Assuming fluxes of this magnitude occurred over 30%-100% the total coastal boundary zone and that fluxes of this magnitude occur all year long, we estimated that 41-138 kg/year Ʃ 5 PCM could be lost to volatilization in Lake Ontario. Lake Erie data yielded an average Ʃ 5 PCM flux of 13 ng/m 2 /day, suggesting that 22-74 kg/year Ʃ 5 PCM could be lost to volatilization in Lake Erie. This may be an overestimate, as fluxes could be lower in the winter, when the surface waters freeze and lower temperatures drive down PCM vapor pressure, but the absence of wintertime dissolved concentration data prohibited flux calculations for these months. Although these estimations are based on temporally-and spatially-102 limited data, they are of a similar magnitude to those estimated in previous Great Lakes studies, and suggest that volatilization may be a significant loss process for dissolved PCMs in this region.

ACKNOWLEDGMENTS
We would like to acknowledge funding from the US EPA Great Lakes Restoration Initiative (GLRI) GLAS #00E00597-0, project officer Todd Nettesheim.
We would also like to express our gratitude to Professor Peter August (URI) for assistance with GIS analysis, David Adelman (URI) for sampler preparation and field deployment logistics, Camilla Teixeira and the field staff of the Emergencies,      1 a N is the number of sites of each type.   The mass of HHCB accumulated in a 2-g PE in response to the simulated air and water concentrations is shown on the left, along with the air-water exchange flux that would be calculated using this pair of air and water PEs.

Scenario)2:)Steadily)Increasing) Air)and)Decreasing)Water) Concentrations
Relative (  Air-water exchange fluxes are shown for shoreline Lake Erie and Lake Ontario sites, as well as nearshore Toronto buoy sites. Positive bars represent volatilization while negative bars represent absorption. Cases where both air and water concentrations were <DL are marked "<DL". Cases where fugacity ratios were not significantly different from equilibrium are marked "X". Offshore Lake Erie and nearshore northern Lake Ontario sites as well as some shoreline sites (SHF, ERI, DUN, BUF, and CV) were omitted because no significant exchange fluxes were calculated there. Error bars represent standard deviation calculated via error propagation. Loss data for benzo(a)pyrene-d12 were not included in determining f for target compounds because loss of this compound was generally greater than loss of pentabromobiphenyl, suggesting that loss due to processes besides PE-air or PE-water diffusive exchange may have occurred. Six air samples were found to have loss of dibromobiphenyl < 90%, which is unrealistic given the deployment times for these 115 samplers and may indicate some inconsistency in deployment practices.

SUPPORTING INFORMATION: POLYCYCLIC MUSKS IN THE AIR AND WATER OF THE LOWER GREAT LAKES: SPATIAL DISTRIBUTION AND VOLATILIZATION FROM SURFACE WATERS
Concentration estimates for these samples were discarded before further analysis and interpretation. (S1) For compounds that were not identical to PRCs, PRC loss data was used to with the exception of δ DBL , which is difficult to observe or measure.
To determine δ DBL for each deployment, PRC loss data for each sample were entered along with compound properties for each PRC: K PEW and K PEA at 298 K, molar volume (V i ), molar mass (M i ), enthalpy of vaporization (ΔH vap ) and internal energy of aqueous dissolution for the sub-cooled liquid (ΔU w ) in kJ/mol and the bestfit δ DBL value was determined using the non-linear least-squares fitting function nls in R, as was used by Booij et al. 5,4 While theoretically, δ DBL could be affected by compound properties, previous studies have reported that variation among compounds is small, 1 so in this study a single δ DBL value was calculated for each set of PRC loss data.
Once values of f were determined for all PRCs, these were used along with known and estimated physico-chemical properties for these PRCs (as listed in Table   S3-8) to calculate the best-fit value of δ DBL for each deployment by using nls in R and and D PE were corrected to the average deployment temperature using data from nearby meteorological buoys (Table S3-2) using Equations S8 -S12. (S11) K PEA,T2 was determined from K PEA via Equation S11, where H vap is the enthalpy of vaporization for the target compound and R is the ideal gas constant.
(S12) K PEW,T2 was determined from K PEW via Equation S12, where U w is the energy of solvation in J/mol for the target compound.
Finally, the concentration measured in pg/kg PE (C PE ) was corrected to the ambient concentration of the target PCM (C a ) using the density of PE (d PE ; 0.91 kg/L), K PEM , and f , as in Equation S13. (S13) Air-Water Exchange Flux Calculations. The equilibrium concentration in ng/g PE (C ∞ ) was determined from C PE as in Equation S14 using f for the target compound. (S14) The fugacity ratio was then calculated as in Equation S15, where C ∞,w and C ∞,a are the estimated equilibrium PE concentrations in water and air samplers. (S15) The air-water exchange flux, F a/w in pg/m 2 /day was calculated using Equation S16 , where v a/w is the mass transfer coefficient, C ∞,w and C ∞,a are the PE equilibrium concentrations, and K PEW,T2 is the PE-water partitioning coefficient corrected for the mean deployment temperature. Mass transfer coefficients were calculated for all PE pairs that displayed fugacity ratios significantly different from equilibrium after error propagation.
(S16) v a/w was calculated as in Equation S17, where K aw is the air-water partitioning coefficient at the mean deployment temperature, v a is the air-side mass transfer velocity, and v w is the water-side mass transfer velocity. (S17) v a was determined from v H2O,a (cm/s), the air-side mass transfer velocity of water in air, scaled for the diffusivity of the target compound in air, D ia (cm 2 /s) at 298 K versus the diffusivity of water in air at 298 K (0.27 cm 2 /s), as in Equation S18.
(S18) v H2O,a was determined as in Equation S19 from the wind speed at 10 meter height, u 10 , as in Schwarzenbach et al. 2003. 7 u 10 was determined from mean wind speed, u, and height of the wind monitor on the meteorological buoy, h, as in Equation S20. In cases where the height of the monitor was not reported, the height was assumed to be 10 meters.
(S19) Elimination System (NPDES) 11 were within 1.5 km of the deployment location.
Though this site was near the mouth of the Oswego River, it is likely that elevated dissolved PCMs were not representative of typical river discharge. Rather, they were likely influenced by these nearby point sources which discharged directly into Lake Ontario.
The second-most elevated dissolved Ʃ 5 PCM was measured at three Toronto nearshore sites (ETOR, TOR, and WTOR), possibly due to influences of three WWTP discharges within 1-8 km of the sites, combined with runoff from the densely populated Toronto conurbation. 12,13 Similarly, Cleveland area sites (CLE, FH, and SHF) were likely influenced by impacted waterways such as the Cuyahoga River and Black River as well as a number of WWTPs that discharged directly into Lake Erie along this stretch of shoreline.
The water sampling site in Rochester (ROC) was placed slightly upstream of the mouth of the Genesee River, making it likely that river discharge was sampled.
This site was likely representative of a mixture of urban runoff and wastewater effluent discharged into the river. In contrast, the PEs at the Buffalo site (BUF), where aqueous PCMs were <15 pg/L, were placed on the mouth of the Buffalo River, but there were no major NPDES-licensed treatment facilities on the river, and this region was characterized by a number of smaller industrial dischargers, which were expected to be much less important as sources of PCMs.
The Niagara River site (NIA) also exhibited elevated dissolved PCM concentrations during late summer. Few NPDES-designated point sources were nearby. However, due to the large volume of discharge and large plume extent from Niagara River, concentrations were expected to be representative of upriver sources channeled into Lake Ontario, including several major WWTPs.         For shoreline/nearshore sites, DEP 1, 2, and 3 are early summer, mid-summer, and late summer/early fall, respectively. For nearshore Northern Ontario sites, DEP 1 is earlyto-mid-summer and DEP 2 is mid-summer to early fall. Concentrations are not available for all deployments at all sites, and the absence of a bar means no sample was retrieved or all concentrations were <DL.

Figure S3-4. Summary of Gaseous PCMs Over Multiple Deployments.
For shoreline sites, DEP 1, 2, and 3 are early summer, mid-summer, and late summer/early fall, respectively. For nearshore Northern Ontario sites, DEP 1 is early-to-mid-summer and DEP 2 is mid-summer to early fall. Concentrations are not available for all deployments at all sites, and the absence of a bar means no sample was retrieved or all concentrations were <DL.     In this study, extracts from PEs deployed in the air and water throughout Lake Erie and Lake Ontario were analyzed for 12 PBDEs and 9 NHFRs to (i) determine baseline concentrations of PBDEs and NHFRs at shoreline, nearshore, and offshore sites, (ii) determine whether the lower Great Lakes were acting as sinks or secondary sources of PBDEs via air-water exchange, (iii) investigate spatial trends of PBDEs and their relation to population centers, and (iv) build a geostatistical interpolation model to provide estimates of dissolved PBDE concentrations across the lakes.   Quality Control. Every batch of PEs was extracted alongside a method blank and two spiked blanks to control for compound losses during extraction, concentration, and cleanup. Average spike recoveries ranged from 67±15% for BDE 2 to 101±19%

Sampler Preparation and
for BDE 100 (Table S4-3). 149 Concentrations were blank-subtracted using the most relevant field blank and detection limits were defined as the upper limit of the 95% confidence interval for 11 laboratory blanks. Detection limits per gram PE are summarized in Table S4-4, and are converted to typical ambient air or water concentrations in Table S4 Posterior distributions for the estimated parameters β 0 , β, σ 2 , and ϕ are shown in   were located near the mouths of tributaries.
Ʃ 12 BDE at offshore sites was generally <3 pg/L, significantly lower than shoreline/nearshore concentrations (p<0.05 two-tailed t test with unequal variance).
The greatest offshore concentrations were observed at the westernmost offshore sites on each lake, with Σ 12 BDE of 2.8 pg/L in western Lake Erie and 3.2 pg/L in western Lake Ontario. These sites were the closest offshore sites to the major rivers feeding each lake (the Detroit River and Niagara River) and may have been influenced by inputs from these rivers.
Generally, dissolved PBDEs in this study were lower than in previous studies. semipermeable membrane devices (SPMDs) and found that total concentrations generally did not exceed 2 pg/L. 9 We investigated whether the discrepancy between concentrations reported by Venier et al. 29 and our own could be due to the presence of dissolved organic carbon (DOC) that was likely cosampled by Venier's active sampling method. As detailed further in the SI (Table S4- NHFRs were detected in more than two aqueous PEs.
Percent detection was low for all NHFRs. As the NHFRs are low-volatility compounds that are expected to be found primarily in the particulate phase, it may be that concentrations in the truly gaseous or dissolved phase were too low to be detected here using passive samplers.

Air-water Exchange of PBDEs.
Fugacity ratios (f w /f a ), which indicate the direction of air-water exchange, are presented in Table S4-13 for all PBDE congeners. In all cases where fugacity ratios were significantly different from equilibrium after error propagation, they indicated absorption into surface waters.
Exchange fluxes for all available air-water PE pairs at each site were averaged to yield mean summer air-water exchange fluxes for each location (Figure 4 24 The occurrence of offshore volatilization in that study, compared to near-equilibrium conditions at offshore sites in this study, may have been due to the smaller surface areas and more urbanized shorelines of Lake Erie and Lake Ontario in comparison with Lake Superior. Liu et al. observed volatilization of polychlorinated biphenyls (PCBs) at the majority of the same sites discussed here, 34 suggesting that the lakes were acting as secondary sources of these legacy pollutants while continuing to absorb PBDEs. Results of this correlation analysis suggest that PBDEs and PCMs share common sources to the aquatic environment. They may also share common sources to the atmosphere, but results for air were inconclusive. Differences in correlation strength between air and water data could be caused in part by the use of slightly different sampling locations for some air versus water PEs), but these differences were not expected to greatly affect correlation strength.

Correlation between PBDE Congeners and Other
Gaseous PBDEs and Population Density. Population data within a 180° wedge to the south of each site resulted in stronger correlation with gaseous Ʃ 12 BDE than population within a circle around each site or population to the north, east, or west.
This was also generally true for individual BDE congeners. Correlations found using a circular radius or 180° southern wedge are compared in Figure S4 Maps of predicted aqueous Σ 12 BDE across Lake Erie and Lake Ontario are presented in Figure 4-5 and variance for these predictions is presented in Figure S4        (S1)

FIGURES AND TABLES
f was estimated using data from performance reference compounds (PRCs) that were loaded into each PE prior to deployment. The initial mass of PRC loaded into the samplers was determined by measuring PRCs in quality control samples (blanks and field blanks), which were prepared alongside those used in field deployments, but never deployed in the environment. f for each PRC was calculated as in Equation S2, where N is the mass of PRC in the deployed PE and N 0 is the mass in the nondeployed blank.
Loss data for benzo(a)pyrene-d 12 were not included in determining f for target compounds because loss of this compound was generally greater than loss of pentabromobiphenyl, suggesting that loss due to processes besides PE-air or PE-water diffusive exchange may have occurred. Six air samples were found to have loss of dibromobiphenyl < 90%, which is unrealistic given the deployment times for these samplers and may indicate some inconsistency in deployment practices.
Concentration estimates for these samples were discarded before further analysis and interpretation. (S2) Values of f for all PRCs were used along with known and estimated physicochemical properties for the PRCs (Table S4- (S13) The unitless fugacity ratio was then calculated as in Equation S14, where C ∞,w and C ∞,a are the estimated equilibrium PE concentrations in water and air samplers, respectively. (S14) The air-water exchange flux (F a/w ) in pg/m 2 /day was calculated as in Equation S15 , where v a/w is the mass transfer coefficient in m/s, C ∞,w and C ∞,a are the estimated PE concentrations at equilibrium in pg/m 3 , and K PEW,T2 is the PE-water partitioning coefficient at the mean deployment temperature. v a/w was calculated for all PE pairs that displayed fugacity ratios significantly different from equilibrium after error propagation. Mean v a/w ranged from 1x10 -7 m/s for BDE 183 to 8x10 -7 m/s for BDE 2.
(S15) v a/w was modeled using a two-film model, much like was done for PE-matrix diffusive exchange. v a/w was calculated in cm/s as in Equation S16, where K aw is the unitless air-water partitioning coefficient at the mean deployment temperature, v a is the air-side mass transfer velocity in cm/s, and v w is the water-side mass transfer velocity in cm/s. . In cases where the height of the monitor was not reported, the height was assumed to be 10 meters.
(S18) (S19) v w,T2 , the water-sid at the mean deployment temperature, was determined from the mass transfer velocity of CO 2 (cm/s) in water (v CO2,w ), scaled to the target compound using the Schmidt number at the deployment temperature (Sc T2 ), the Schmidt number for CO 2 at 20°C (Sc CO2,w = 600) and a sc , a scaling factor based on wind speed (0.67 for mean wind speeds <4.2 m/s, and 0.5 for higher wind speeds).
The dimensionless Schmidt number is the ratio of the viscosity of the water to the diffusivity of the target compound in water.
(S20) v CO2 , the average velocity of CO 2 (cm/s) in water over the deployment, was determined by integrating the Weibull probability density function using all recorded wind speeds over the deployment from the nearest meteorological buoy, scaled to 10 m above the interface as in Equation S19.

Most buoy data was accessed from online databases provided by Environment Canada and the National Data Buoy Center (NDBC). In some cases, data from temperature loggers maintained by the Ontario Ministry of the Environment and
Climate Change from the deployment buoys were available (OME Logger).      [27][28][29] Results suggested that the discrepancies between the studies could not be entirely explained by sorption to DOC alone, as concentrations of DOC much greater than 3 mg/L would be needed in most cases, and the samples being compared were from open-lake and nearshore sites, rather than shoreline sites.
(S28) (S29)      and north-south components (u ave and v ave , respectively). This data is also summarized in Figure S4-5, which is a map with approximate average direction at each buoy site marked by an arrow.

PBDEs in Air (pg/m 3 ): Strongest Correlation with Population within a Circular Area (left) and with Population in a 180º Wedge South (right)
To calculate average wind direction, direction in degrees was broken down into east-west and north-south components and then each component was averaged over the deployment period:       The gaseous fraction of ambient air has a distinct composition compared to the particle-bound fraction. 6 The total amount of PAHs in the gaseous phase is generally greater than in the particulate phase, though total PAHs in this phase are dominated by as personal monitoring devices for exposure to gas-phase HOCs and frequently detected several 2-3-ring PAHs, as well as some OPEs. 14 Chronic exposure to gas-phase OPEs and PAHs in ambient air is of concern because these compounds have been associated with carcinogenicity, endocrine disruption, and other biological effects in previous in vitro and in vivo studies. 6,[15][16][17][18] Activation of the aryl hydrocarbon receptor (AhR) is linked to induction and repression of a large number of genes, modulation of cell growth and proliferation, tumor promotion, immunological effects, cardiotoxicity, and endocrine disruption, with the severity and type of response dependent upon the specific ligand and its binding affinity. 19 Many previous studies have assessed health risks of ambient air pollution using induction equivalency factors (IEFs) to represent the relative AhRmediated potency of PAHs relative to benzo(a)pyrene (BaP). 8,20 This approach assumes an additive, rather than synergistic or antagonistic, relationship between multiple ligands. AhR is activated by binding with variable affinity to several PAHs, with 4-5-ring PAHs generally more potent than the 2-3-ring PAHs that dominate gasphase air pollution. 6 Highly carcinogenic PAHs such as benzo(a)pyrene (BaP) are typically present only at very low concentrations in the gas phase due to low volatility. The lower molecular weight PAHs, especially phenanthrene, fluoranthene, and the methylated phenanthrenes/anthracenes, are expected to contribute more significantly to the potency of this fraction due to their high gas-phase concentrations. 6  Many previous studies have noted that gaseous HOCs should not be ignored in risk assessments, but they were all carried out using high-volume air samplers or passive polyurethane foam (PUF) samplers, which are less selective for gaseous HOCs than diffusive uptake by polyethylene. 23 This study is the first to investigate AhR activation caused by the freely gaseous fraction of HOCs taken up by a single-phase sampler consisting only of pre-cleaned polyethylene, and will help contribute to our knowledge of the biological relevance of the truly gaseous fraction of ambient air.
Passive samplers of this type have similar affinity for HOCs as organism lipids, and have been used in predicting the extent to which HOCs will bioaccumulate. 24 The composition of HOCs taken up into the polyethylene matrix is similar to the composition that would be found in biological tissue.
Polyethylene passive samplers (PEs) were deployed throughout the Cleveland (OH) area on the southern shore of Lake Erie during June-September, 2013. Extracts from PEs were analyzed for PAHs and OPEs and were also analyzed via an in vitro bioassay to measure AhR induction. A map of the study region is shown in Figure 5-1 and characteristics of the deployment sites are summarized in Table 5 (Table S5-1).
To avoid interference with biological assays, samples were not spiked with internal standard prior to extraction and so were not corrected for internal standard recoveries. Dosing solution concentrations were not blank-subtracted before use in data interpretation. This was considered appropriate as we were primarily interested in determining the actual concentration present in the bioassay exposure solution. f (x) = d c 1 + exp(b(log(e) log(x)) Eq 1 In addition to the EC 50 , EC BaP20 and EC BaP50 were calculated as alternative measures of relative potency. The EC BaP20 and EC BaP50 are the doses achieving 20% and 50% of the effect observed for the positive control, 1.2x10 -7 M BaP. The EC BaP50 was identified as a more useful metric than EC 50 because the extracts' dose-response curves were not parallel and maximum efficacy varied among curves.

Ambient Air
Dosing solutions were prepared so that each sample was representative of the same amount of extracted PE to facilitate comparison with the PE blank. However, due to site-to-site variability in sampling rates, the volume of air represented by each sample was different (Table 5- BaP Eq chem = ⌃(IEF n · C n )(ng/µL) Eq 2 For comparison to bioassay results, the relative potency of each sample extract was expressed as the amount of BaP that would be needed to achieve the same response. The bio-derived toxic equivalency (BaPEq bio ) was calculated as in Equation  (Table S5-2), OPEs (Table S5- OPEs generally varied independently of one another, though some degree of correlation (r 2 > 0.3) was observed between some pairs, including TCEP and TCIPP, and TDCIPP and EHDPP (Table S5-6). There were few correlations found between the PAHs and OPEs, though TnBP was found to correlate with 2-4-ring PAHs (Table   S5-7).  (Table 5-2). For this reason, EC BaP50 , measured relative to the plate-specific positive control, was used to compare potency of samples.
The EC BaP50 and EC BaP20 of each extract, normalized for volume of air sampled during each deployment, are displayed in Table 5 The three rural/residential sites had the lowest potency (greatest EC BaP50 values), ranging from 2.6 -6.6 g PE/mL, followed by the two Cleveland Lakefront 238 sites. The most potent extracts were from the three Cleveland Downtown sites and one semi-urban residential site (University Heights, a densely populated suburb). This contrasts with work by Klein et al., where no change in potency of gaseous extracts was observed between urban and rural samples with distinct chemical compositions, but is consistent with work by Ersekova et al, where extracts from impacted sites were found to be more potent in AhR bioassays than extracts from rural sites. 10,22 The relative potency of the PE Blank (EC BaP50 = 23±5 g PE/mL) was significantly lower than all field samples when compared using the EC BaP50 values, prior to adjusting for volume of air sampled. Blank comparisons were done before normalizing for the volume of air sampled so that each sample would be representative of the same mass of extracted polyethylene.
Relative potency and maximum efficacy of the extracts did not appear to be correlated. This is most likely due to a complex interplay between the unique composition of ligands in each sample, their affinity for the AhR, the resulting ligand- observed steadily decreasing potency in extracts from 24 to 72 hpd. 20 This difference is most likely due to differences in induction kinetics and increased stability of the GFP reporters compared to the luciferase reporter. 31 It is also possible that some of the response observed in this study was due to compounds that were less readily metabolized than PAHs and OPEs.

Bioassay-Derived BaP Equivalents for PE Extracts. A map displaying results for
BaPEq bio is displayed alongside maps of total concentrations of PAHs and OPEs in the dosing solution (Σ 40 PAH and Σ 12 OPE) in Figure 5 (Table S5-8).
BaPEq chem values calculated using potencies from Machala et al. ranged from 1.6 to 7.9 ng/µL BaP, as shown in Table 5

CONCLUSIONS
This study demonstrated the use of PEs coupled with in vitro bioassays as an approach to measure cumulative biological effects of ambient gaseous air pollution.
While some activity was seen in the PE blank, the relative potency of field samples was found to be significantly elevated above blank levels, suggesting that interference from the PE matrix or typical laboratory contamination did not prohibit the use of PE extracts in bioassays for AhR-mediated potency.

INTRODUCTION
Polyethylene passive samplers (PEs) and other passive samplers, including semi-permeable membrane devices (SPMDs) and polyurethane foam disks (PUFs), are an increasingly popular option for air pollutant monitoring projects. Passive samplers have been used to analyze spatial trends and identify probable sources of emerging and legacy semi-volatile organic contaminants (SVOCs) on global, regional, and citywide scales, [1][2][3][4][5] and to analyze seasonal and long-term temporal trends in concentrations of gaseous SVOCs. 6,7 They are also being deployed indoors and worn on the body to assess health risks associated with occupational exposures in workplaces and homes. [8][9][10] Passive samplers simultaneously deployed in different media are becoming an increasingly popular tool for measuring fluxes of SVOCs, including air-water diffusive fluxes. [11][12][13] PEs accumulate SVOCs from air passively, selecting for the non-particlebound portion of ambient air contaminants. PEs and other passive samplers are promising tools because they are cost effective and simple to deploy; they accumulate SVOCs via diffusion, so they require no power source to operate, enabling long-term monitoring at unprecedented high resolution and in remote locations. They tend to have lower detection limits than active samplers, allowing quantification of trace-level contaminants in the gaseous phase that would be challenging to detect using a traditional sampling apparatus on feasible timescales.
Despite ease of use, there are some drawbacks to choosing passive samplers rather than traditional active samplers for a monitoring project. Unlike active samplers, the rate at which air is sampled by passive samplers cannot be set at a constant value. One main concern inherent in the use of passive samplers is accurately estimating sampling rates so that concentrations in the sampling matrix can be dependably converted to ambient air concentrations. This is especially challenging for emerging contaminants with poorly constrained chemical properties.
PEs are usually contained with protective housing, and so transfer is thought of as a three-step process by which (i) air is transferred from the surroundings to within the sampler housing, (ii) air is transported to the PE-air interface, and (iii) exchange occurs at the PE-air interface. 14 At the PE-air interface, turbulent mixing becomes less important and the exchange process is dominated by molecular diffusion. The The model for uptake of gaseous compounds into the passive sampler matrix is based on Whitman Two-Film Theory, originally developed to describe the "driving potential" behind absorption of a gaseous solute into the liquid phase, which was known to be proportional to the "distance from equilibrium". 17     would depend on other estimated properties, rather than on direct measurement. Any K PEA calculated for a non-equilibrated compound were "lower-bound" estimates, becoming further and further from the true value as distance from equilibrium increased.  PRC Loss in PEs. PRC loss data is shown in Figure S7  PE uptake profiles for many equilibrated compounds likely responded to this pulse, as 320 concentrations in PEs increased initially, and then began to decrease later in the study.

Summary of HV-AAS
Values of K PEA calculated for the majority of equilibrated compounds based on activeand passive-derived data from this study were lower than those derived from vapor pressure or previous empirical measurement, though the reason for this discrepancy is unknown.        TnBP  tri--n --butyl phosphate  126--73--8  TCEP  tris(2--chloroethyl)         contaminant of emerging concern with unknown impacts on remote marine environments.  Concentrations of OPEs in PE blanks are shown in Table S8 Table S8-3). Based on these results, a δ WBL of 60 µm was assumed for all Arctic surface water PEs (PRC loss data not available). The δ WBL value was plugged into Eq 1 along with the relevant physico-chemical properties for each target OPE to determine the percent equilibration reached during the deployment. Percent equilibration predicted for each OPE at each site is presented in Tables S8-4 for deep mooring deployments and S8-5 for surface water deployments. The five smallest OPEs (V m < 290 cm 3 /mol; log K PEW < 5) were generally estimated to reach >90% equilibrium during deployment (Table S8-4). Among the non-equilibrated compounds, sampling rates calculated from best-fit r δ WBL ranged from about 1-10 L/day at deep mooring sites. Mean current velocity for all PEs at all depths ranged from 7 -13 cm/s. The greatest sampling rates and fastest current velocities were observed at the shallowest deep mooring sites. The lowest sampling rates and slowest current velocities were both found at the two deepest sites in the Eastern Fram Strait.

INTRODUCTION
Boundary layer thickness is plotted with depth in Figure S8-1.
Physico-Chemical Properties. One of the major challenges in understanding the transport and fate of OPEs is the paucity of data on their physico-chemical properties. 8 Here, PE-water partitioning coefficients (K PEW ) were estimated from subcooled liquid aqueous solubility (log C w,sat (L); mol/m 3 ) as in Lohmann 2012. 9  However, more work needs to be done to confirm empirical OPE K PEW values.
Physico-chemical properties used to calculate ambient concentrations are presented in Table S8-6. Values of C w,sat (L) used in K PEW calculations were taken from a collection of estimated properties by Zhang et al. 10 Values calculated from EPI Suite WSKOWWIN were used because this model performed best in predicting C w,sat (L) for a wide range of compounds. 10 However, these C w,sat (L) values were often biased low, meaning that ambient concentrations estimated using these values may also be underestimated. To interpret results for PEs deployed in seawater, values of C w,sat (L) were corrected for salinity, as described above for the PRCs. 7 All K PEW s were also corrected for mean deployment temperature as previously described, 9 assuming an energy of solvation of 25 kJ/mol. As mentioned previously by Booij et al., 3 the effect of pressure in deep ocean regions on physico-chemical properties of organic chemicals is not well characterized, so parameters were not adjusted for pressure effects.
The molecular diffusivity of each OPE in water was calculated for each deployment's mean temperature using the Wilke-Chang equation with 2.6 as the association parameter for water and determining viscosity of water using a table provided by Schwarzenbach et al. 7 Molecular diffusivity in polyethylene was taken from Pintado-Herrera et al. 11 when experimental values were available, and was otherwise calculated from molar volume as in Lohmann 2012, 9 and subsequently corrected for mean deployment temperature using the Arrhenius equation, assuming an activation energy of 100 kJ/mol.  expected that alkyl/aryl-OPEs would be present at lower concentrations than Cl-OPEs, as they have been shown to be more readily degradable than Cl-OPEs by numerous routes, including biodegradation and hydrolysis. 14,15 Surface water concentrations of dissolved Σ 7 Alkyl/aryl-OPE are shown in For the most part, little variation in dissolved OPE concentration was observed with depth. These flat depth profiles could be due to a high degree of vertical mixing, or the release of particle-bound contaminants to the dissolved phase with depth. 17 Booij et al. 3

Concentrations Derived from SPARC C W Estimation
CONCLUSION Spatial Distribution of HOCs. The spatial distribution of gaseous PAHs, PBDEs, and PCMs was found to be influenced by nearby population density. The influence of population centers on ambient concentrations of air pollutants is a significant consideration to regional and global modeling studies, as well as predictions of human health risk. While previous studies demonstrated correlations between concentrations of HOCs in air and population within 25 km, 1-3 here the effect of compound vapor pressure on spatial distribution was explored by using two additional metrics: the radii of site characterization at which maximum correlation was seen, and the maximum distance at which a statistically significant correlation with population remains.
Results implied that more volatile compounds were influenced by more distant population, while less volatile compounds were influenced only by local population.
Questions remain as to how concentrations of gas-phase HOCs change in response to differences in the levels and composition of ambient aerosol from site to site, a consideration that could only be addressed if particulate samples were collected alongside freely gaseous samples. Gustafson et al. hypothesized that HOCs traveling from urban to rural areas may re-condense on background aerosol, causing pronounced urban-rural gradients in gas-phase HOCs. 4 The spatial distribution of particle-bound HOCs with respect to population centers, and its interaction with spatial distributions of gas-phase HOCs, is an important missing piece of the puzzle.
Additionally, investigating the role of atmospheric lifetime differences among different gas-phase HOCs, and the role of aerosols in prolonging these lifetimes by retarding degradation, 5 would add further nuance to the observations made here.
Future work should also include exploration of how the relationship between population and gas-phase HOC concentrations changes seasonally depending on air temperature, which can greatly affect the partitioning of many HOCs between the gasand particle-phase. Additionally, this study should be repeated in other areas of the world and on other scales, as choice of site and resolution of data may factor in to the observations made.
Diffusive Air-Water Exchange of HOCs. PAH air-water exchange was found to vary by compound and deployment period, while PCMs were lost from Great Lakes surface waters via volatilization and PBDEs were absorbed. More work needs to be done in the Great Lakes region to quantify other inputs and losses of dissolved HOCs in order to understand the role played by diffusive air-water exchange in the lakes, and to construct budgets for these toxic pollutants. Additionally, the importance of DOC sorption, gas-phase photodegradation, degradation in surface waters, and settling processes in influencing air-water diffusive exchange, and how these processes change seasonally, require further investigation to better understand observations in this study.
Total losses of PCMs were estimated in Chapter 3 by averaging volatilization fluxes over the coastal boundary zone, arriving at loss estimates for Ʃ 5 PCM of 41-138 kg/yr for Lake Ontario and 22-74 kg-yr for Lake Erie. Loss and input estimates of PCMs can also be calculated by interpolating dissolved PCM concentrations across the lakes using human population as an auxiliary variable (Figure 9-1), and then extrapolating fluxes across the lake based on the correlation between volatilization fluxes of PCMs and the dissolved concentrations that drive them (Figure 9-2). This approach led to similar loss estimates for Ʃ 5 PCM, amounting to 150±230 kg/yr from Lake Ontario and 120±380 kg/yr from Lake Erie. A similar approach was taken to calculate total inputs of Ʃ 12 BDE due to absorption using data from Chapter 4.
Gaseous Ʃ 12 BDE over the lakes was interpolated (Figure 9-3) and used to estimate absorptive fluxes using the regression with gas-phase concentrations shown in Figure   9