AIR-WATER EXCHANGE AND TRENDS OF PERSISTENT BIOACCUMULATIVE TOXICS (PBTs) ACROSS LAKE SUPERIOR

Persistent bioaccumulative toxics (PBTs) have long been studied in Lake Superior air, water, sediment, and biota, however, sampling has often been constrained to a few study sites. Polyethylene passive samplers (PEs) provide the ability to collect time-averaged, tandem air and water measurements from any accessible point along the shore or in open water. PEs were deployed in the air and water at 19 sites across Lake Superior from June-October 2011. Samples were solvent extracted and analyzed using gas chromatography/tandem-mass spectroscopy to quantify 22 polycyclic aromatic hydrocarbons (PAHs), 11 polybrominated diphenyl ethers (PBDEs), 24 organochlorine pesticides (OCPs), and 18 polychlorinated biphenyls (PCBs). PAHs are still emitted to the atmosphere from both anthropogenic activity and natural sources, and flux rates suggest net deposition into Lake Superior from the atmosphere. They were found at most sites, however, distributions appear to be associated with populated and industrialized areas (air and water ranges below detection, bd, to 190 ng/m and bd-130 ng/L, respectively). Retene, a signal of conifer trees and used as an indicator of forest fires, was present across the lake, representing a large portion of total PAH concentrations at rural areas, particularly in the water. PBDEs were similarly connected to developed areas, but were present at much lower concentrations (air and water range bd-19 pg/m and 0.051-7.6 pg/L, respectively). PBDE flux rates were dominated by BDE-47, however, flux direction was not consistent across the lake and deployment periods. PCB concentrations were also greatest near developed areas (up to 57 pg/m in Ontonagon air and 45 pg/L in Sault Saint Marie water), but a ban on direct emissions since the 1970s has resulted in upward flux rates. This net volatilization suggests Lake Superior is acting as a secondary source of these compounds to the atmosphere. OCPs varied greatly between individual compounds, but were overall detected at every site across Lake Superior. Long-banned hexachlorobenzene (range bd-180 pg/m) and α-hexachlorocyclohexane (bd-640 pg/L) dominated the air and water concentrations, respectively. Both exhibited fluxes indicative of near-equilibrium or net volatilization. Conversely, recently-used endosulfan I had negative flux rates across the lake due to ongoing terrestrial emissions. Continuing deposition trends of currently used PBTs, changes in recently-banned PBTs, and possible secondary emissions of long-banned PBTs must continue to be monitored across Lake Superior and the other Great Lakes in order to evaluate the efficacy of regulatory measures and threats to human and environmental health. PEs provide an easy and affordable alternative to active sampling ideal for high resolution sampling of large regions, such as Lake Superior.

The Integrated Atmospheric Deposition Network (IADN) monitors trends of nonpoint source pollution in the Great Lakes basin (Buehler and Hites 2002). IADN is a joint effort between Canada and the United States established in 1990 to determine whether the concentrations of PBTs in air and precipitation near the Great Lakes are changing as a function of time (Hillery et al. 1998;Buehler and Hites 2002). Five master stations, one on each lake, were selected to represent regional air near the lakes with minimal impact from local sources (Hillery et al. 1998). Lake Superior's master station, Eagle Harbor, MI, has concentrations among the lowest measured (Buehler et al. 2001). As the least industrialized and least populated of the Great Lakes (Hillery et al. 1998), Lake Superior air quality is considered to be representative of regional background atmospheric concentrations.
Lake Superior is the largest freshwater lake in the world by surface area (82,100 km 2 ) and third largest by volume (MN Sea Grant 2012). It holds more water than the other Laurentian Great Lakes combined (MN Sea Grant 2012), accounting for 28% of the world's freshwater supply (Buehler and Hites 2002). Lake Superior has a relatively small watershed (127,700 km 2 ) containing 848 tributaries ( Despite its rural location, water and air quality in the Lake Superior region are affected by anthropogenically produced PBTs. PBTs are semivolatile compounds making them available for atmospheric transportation in both the gas and particle phases (Gevao et al. 1998). Thus, through a repeating process of evaporation and precipitation, these persistent pollutants can be distributed to remote areas far from known sources (Gustafson and Dickhut 1997;Baker and Eisenreich 1990). The small, rural watershed and large surface area (catchment:surface area 0.6:1) mean that atmospheric deposition is the major input pathway for PBTs to Lake Superior. Net atmospheric loadings to Lake Superior are controlled by five main factors: atmospheric-tributary inputs, wet deposition, dry deposition, gas transfer, and bubblespray production (Gustafson and Dickhut 1997;Hillery et al. 1998). In addition to the influence of PBT point sources, atmospheric concentrations of semivolatile organic compounds have a seasonal component whereby ambient temperature determines the direction of PBT transfer across the air-water interface (Hillery et al. 1998). Particularly in regions with lower PBT concentrations, cold temperatures in the winter and spring can lead to net PBT deposition, whereas peak temperatures in the fall may lead to net volatilization ; Morgan and Lohmann 2008;Sabin et al. 2010).
Temperature control on atmospheric PBT concentrations and air-water exchange may become even more significant with climate change in the Great Lakes region. In general, Lake Superior exhibits water stratification characteristics typical of mid-latitude lakes: strong positive stratification during the summer, one or more isothermal mixing events during spring warming and fall cooling, and seasonal ice formation causing weakly negative stratification in the winter (Austin and Colman 2007; Jay Austin, University of Minnesota, personal communication). However, Lake Superior is one of the most rapidly warming lakes in the world (LSBF 2013), altering its seasonality. Since 1980, surface water temperature in the summer has increased ca.
1 o C per decade and regional air temperature has increased by 0. days per year, earlier summer stratification, and a deeper warmed layer (Moen 2008).
Higher temperatures could mean an increase in evaporation of surface water of 7-17% by 2030 and more intense rain, worsening wastewater overflows (Moen 2008).
Combined with drought and too little snow to feed the lakes, these factors have resulted in water levels in the Great Lakes below their long term averages for the past 14 years and historically low water marks (NYT 2013). In the Great Lakes Water Quality Agreement, the U.S. and Canada recognize that climate change may affect the use, release, transport and fate of chemicals in the Great Lakes Basin Ecosystem, thereby contributing to impacts on human health and the environment (GLWQA

2012). As a result of changing concentrations and increasing evaporation, Lake
Superior may experience greater net volatilization of PBTs and become a significant secondary contributor of pollutants back to the atmosphere.
While IADN's active atmospheric monitoring is useful for tracking emissions trends and can indicate deposition rates, there are several limitations which render high-volume air sampling unsuitable for regional monitoring of PBT distributions.
Active samplers are noisy and require electricity and maintenance (Khairy and Lohmann 2012). Sampling is not continuous; it consists of 24 hours of high volume (hi-vol) air collection every 12 days to assess year-round spatial and temporal trends of organic pollutants in gas, particle, and precipitation phases (EPA 2012;Hillery et al. 1998;Buehler and Hites 2002). High volume air samplers use filters, which can experience breakthrough and degassing of pollutants from particles and can affect the measurement quality (Khairy and Lohmann 2012). Additionally, the IADN active air sampling program excludes water concentrations and contributions to the atmosphere as a result of volatilization, affecting the assessment of net fluxes across the air-water interface. As an alternative, polyethylene (PE) passive samplers can be independently deployed for days or weeks at a time to establish time-averaged concentrations of trace compounds and do not require a power source. This independence means there are no limitations on spatial distribution, permitting higher resolution monitoring of large geographic areas. Comparisons between PEs and active sampling find an agreement generally within a factor of 2 (Khairy and Lohmann 2012;Khairy and Lohmann 2013).
Passive sampling is a widely accepted method for measuring active (truly dissolved or vapor-phase) compounds in water and atmosphere (Lohmann 2012; Morgan and Lohmann 2008;Lohmann et al. 2012). Compared to other passive sampler matrices, PE is the simplest in chemical makeup and the cheapest polymer available (Lohmann 2012). Low-density polyethylene (LDPE) is produced by free radical polymerization and is made up of ca. 20-40 long and short chain branches per 1000 carbon atoms off the main PE chain (Lohmann 2012). LDPE density is generally between 0.91-0.923 g/cm 3 . Simple commercial 2 mil dropcloth can be easily obtained and prepared.
PE passively accumulates hydrophobic organic compounds in proportion to their active concentrations, which is directly related to their availability for mass transfers and bio-uptake in the environment (Adams et al. 2007;Lohmann et al. 2012).
PBTs are nonpolar or weakly polar, hydrophobic substances which are largely sorbed to particulate material, meaning their freely dissolved concentrations are extremely low . PE absorbs hydrophobic compounds from the aqueous and gaseous phases and concentrates them from their trace levels ). Cavities in the polymer matrix are only on the order of 1 nm, reducing the uptake of particle bound compounds and limiting sampling to gas and dissolved phase compounds .
Hydrophobic organic contaminant uptake is described by the partitioning constant between PE and water or air, KPEx, at equilibrium: where CPE is the concentration of the compound in the PE (mol/kg PE) and Cx is the concentration of the compound freely dissolved in water (mol/L H20) (Lohmann 2012) or gaseous in the atmosphere (mol/L air). The magnitude of KPEw is almost entirely dominated by a compound's solubility in water and we can assume constant solubility in LDPE (Lohmann 2012). Uptake of hydrophobic compounds is an absorptive process whereas surface adsorption is not significant (Lohmann 2012).
The objective of this project is to enhance our understanding of the distribution of PBTs across Lake Superior in order to evaluate the effects of local and long-range inputs. The GLC funded this project to achieve two main goals: (1) uniquely enhance measurements of the spatial variability of atmospheric concentration of PBTs around Lake Superior, and (2) assess whether the lake is volatilizing or absorbing gas-phase PBTs to derive fluxes and loading to Lake Superior.
Passive samplers, such as PEs, are ideal for sampling PBTs across Lake Superior and the Great Lakes region in general. In the following study, we employ PE passive samplers to collect air and water samples across the entire surface of Lake Superior. Passive samplers are inexpensive, light-weight, and do not require a power source, so they can be deployed by volunteers at any dock around the perimeter and on any buoy in the open lake. Currently, IADN samplers are only operating at two locations (Brule River and Eagle Harbor), limiting the scope of influence from point sources around the lake perimeter. Expanded sampling coverage of Lake Superior can significantly improve the spatial resolution of PBT concentrations in both the air and water. PEs can also be deployed in quick succession, providing a continuous, timeaveraged dataset to monitor changes in PBT concentrations with time.
Coastal population distributions and changes in PBT production based upon enforced policies can affect PBT trends and should be tracked in Lake Superior, especially if we depend upon those atmospheric concentrations to determine background conditions in the Great Lakes region. Increasing emissions of emerging contaminants will lead to deposition into Lake Superior, whereas continued decreases of banned PBTs in the atmosphere could cause an equilibrium shift across the airwater interface whereby the lake may become a secondary source of toxic compounds back to the atmosphere. Additional stressors related to climate change underscore the importance of expanded and continued monitoring of PBTs in the Lake Superior region.

Introduction
Polycyclic aromatic hydrocarbons (PAHs) and polybrominated diphenyl ethers PAHs are composed of two or more fused aromatic rings pyrogenically released as the result of incomplete combustion of both anthropogenic (fossil fuels) and natural (biomass) sources (Gewurtz et al. 2008;Slater et al. 2013). They can also result from petrogenic releases (oil spills) and volatilization from polluted grounds (Sabin et al. 2010;Khairy and Lohmann 2012). PAHs are actively emitted to the environment through car exhaust, industrial activity, residential heating, and wildfires (Buehler et al. 2001), leading to a ubiquitous presence and continued atmospheric deposition across the Great Lakes (Slater et al. 2013).
Likewise, PBDEs are emitted from industrial and urban centers where they are incorporated in a wide range of products, from textiles and polyurethane foam to circuit boards and are subsequently transported and deposited by the atmosphere (Crimmins et al. 2012 (Crimmins et al. 2012). PBDEs are still in imported products, suggesting continued release to the environment and deposition into Lake Superior.
The objective of this study is to enhance our understanding of the distribution and behavior of PAHs and PBDEs across Lake Superior. We deployed PEs in the air and water at 19 sites across Lake Superior from April-October 2011 with the aim to (1) enhance the spatial coverage of air and water sampling stations across Lake Superior; (2) determine PBT distributions, concentrations, and air-water exchange; (3) discern seasonal temporal trends; and (4) evaluate the efficacy of PEs as tools to monitor regional PBT distributions.

Sampling Methodology
Low density polyethylene (2 mil Standard checks were analyzed every ten samples to monitor instrument performance.

Calculations/Data Analysis
Total Σ22PAH concentrations were determined for each site. The 22 PAHs are naphthalene, biphenyl, acenaphthylene, acenaphthene, fluorene, phenanthrene, anthracene, methyl phenanthrenes, fluoranthene, pyrene, retene, benz(a)anthracene, chrysene, benzo(b)fluoranthene, benzo(k)fluoranthene, benzo(a)pyrene, benzo(e)pyrene, perylene, benzo(j)fluoranthene, indeno(1,2,3-cd)pyrene, dibenzo(a,h)anthracene, benzo(g,h,i)perylene. Sums of PBDEs were also determined for each site from the congeners 8,15,28,30,47,49,99,100,153,154. Truly dissolved concentrations of PBTs in water, CW (ng/L), were calculated according to Adams et al. (2007): CeqW is the concentration of the PBT in the PE when in equilibrium with water (pg/kgPE), and KPEW is the PBT partitioning coefficient from water into polyethylene (L/kg). KPEW values were obtained from Lohmann (2012) and temperature-corrected according to a modified form of the Van't Hoff equation where ΔHvap is the enthalpy of vaporization (kJ/mol), R is the gas constant (kJ/K*mol) and T is the average water temperature during deployment (Kelvin where Rs is the sampling rate used to adjust the curve (L/day), t is the sampling time (days), and V is the PE volume. Performance reference compound equilibrium was calculated by comparing the concentrations remaining at time t to the concentrations at time 0 (assumed to be equivalent to concentrations in the field blanks).
The gradient of air-water exchange is the ratio of the equilibrium concentration of the PBT in air (CeqA, pg/L) to the equilibrium concentrations of the PBT in water (CeqW, pg/L), where a gradient >1 indicates atmospheric deposition and a gradient <1 indicates volatilization from the water to the air. Net air-water fluxes (ng/m 2 /day) were calculated by modifying the equation Flux = kol*(CW-CA/KAW) (Schwarzenbach et al. 2003), to those studies did not include Sault Saint Marie. Although not the largest city along the Lake Superior shoreline, Sault Saint Marie is located on the Soo Locks, which sees >70,000,000 net tons of cargo annually (USACE 2011 (Hafner and Hites 2003).
Relative contributions from petroleum and combustion can be assessed by establishing characteristic ratios of PAHs, such as anthracene/anthracene+phenanthrene, fluoranthene/fluoranthene+pyrene, and benz(a)anthracene/benz(a)anthracene + chrysene ratios (Yunker et al. 2002). The Ant/Ant+Phen ratio at Sault Saint Marie was <0.10 (0.03), which indicated a greater contribution from petroleum. The Flra/Flra+Pyr was just above equilibrium (0.5) at a ratio of 0.69, likely indicating a combination of combustion and petroleum. The less-commonly used BaA/BaA+Chry was <0.2 (0.05), indicating the presence of crude oil and confirming petroleum is a major source atmospheric PAHs in the area.
The second greatest atmospheric PAH concentrations were at Ashland, WI (June-October average 31 ng/m 3 ). Located along the southeastern shore of Lake Superior, Ashland is a major industrial port with a history of iron ore processing, This may be because the PE was deployed at the regional EPA office northeast (upwind) of the city center. It was followed by Marquette (June-August average 10 ng/m 3 ) and Thunder Bay/Welcome Isle (June-October average 5.9 ng/m 3 ).
Concentrations were characterized by many of the same petroleum-and biomass-  (Sun et al. 2006). The site is used as the background atmosphere standard for Lake Superior and the entire Great Lakes region (Cortes et al. 2000 (Sun et al. 2006).
Additionally, PEs only sample the truly gas phase fraction of atmospheric PAHs.
High volume air samplers can yield overestimates of gas PBTs as a result of compounds desorbing from particles onto the filter. Likewise, subsequent entrapment of analytes from the filter can lead to underestimates of gas-phase PAHs, overall making measurements difficult to interpret (Perlinger et al. 2005).  (Sun et al. 2006). Retene can be produced in urban areas from municipal waste incinerators, tire burning, and incineration of building waste; however, it is primarily emitted by burning wood. Rural households consume about three times more wood than urban homes (Sun et al. 2006). It is therefore not surprising that retene accounts for 20-50% of the total PAHs at the northern coastal sites and is the only PAH present in open lake waters. Retene concentrations peak in the winter months when wood is burned for residential heating, but lower-magnitude spikes can occur between April and September due to naturally occurring wildfires (Sun et al. 2006).

Temporal Distributions: Atmospheric temperature nearly doubled between
April-June and remained roughly the same through October, while water temperature  Flux direction = CeqA/CeqW (5) A ratio >1 indicates a higher equilibrium concentration in the air, and therefore, net deposition. A ratio <1 suggests a higher equilibrium concentration in the water, resulting in net volatilization. Significant departure from equilibrium between the two reservoirs was considered to be ca. 290%. Generally, PAHs were deposited into Lake Sturgeon Bay, where retene concentrations in the water were exceptionally high and significant air-water exchange was near equilibrium. Elevated retene was likely due to the proximity of those sites to the Nipigon River, which supplies terrestrial water from the rural Ontario region north of Lake Superior. As previously discussed, retene inputs may have been elevated due to high wildfire activity in the region that season, resulting in net volatilization of retene into the atmosphere at the central open lake and Sturgeon Bay. This volatilization was in contrast to the eastern and western open lake site where retene was being deposited.
Atmospheric deposition is a major source of pollutants to the Great Lakes (Buckley et al. 2004;Li et al. 2006;Gewurtz et al. 2008). Chemical exchange across the air-water interface is one of the processes controlling concentrations and residence times of pollutants in these water bodies (Baker and Eisenreich 1990;Gevao et al. 1998;Hillery et al. 1998). PE sampling improves upon previous estimates of air-water PAH exchange. Simultaneous time-averaged concentrations for both the air and water dampen diurnal effects, minimize anomalous signals, and only equilibrate with gas-or dissolved-phase compounds directly available for exchange across the air-water interface. However, there are uncertainties in gas-transfer loadings estimated by this method. We used the two-film model for air-water gas exchange, which assumes diffusive flux of compounds between the air and water is limited by transport across two thin films at the interface and not by turbulent transport in the bulk mixed layers (Gevao et al. 1998;Perlinger et al. 2005). The rate of gas exchange is given by equation 3 and depends upon the Henry's Law constant, wind speed, and temperature.
All of these parameters have a degree of uncertainty and thereby limit our ability to calculate accurate net fluxes of gas-phase PAHs (see Supporting Information).

Polybrominated Diphenyl Ethers (PBDEs)
Spatial Distributions: Σ11BDE concentrations were more than three orders of magnitude lower than PAHs across Lake Superior (Figure 2.5A). Atmospheric PBDE concentrations were greatest at Marquette (June-October average 13 pg/m 3 ) ( PBDE Sources: PBDE concentrations in both the air and water were correlated with industrial and urban sites, reflecting their origin as flame retardants used widely in commercial and industrial products. Atmospheric deposition is the dominant source of PBDEs to Lake Superior and is the main method of input to midlake sites (Li et al. 2006). The global fractionation theory predicts that lighter congeners will travel farther from PBDE sources (Li et al. 2006). Tri-, tetra-, and pentaBDEs were most commonly detected at Lake Superior sites, whereas lighter and heavier congers were typically below detection. We cannot determine whether this is a reflection of global fractionation of BDEs, their use, or our sampling approach.
Commercial pentaBDE was the most commonly used technical mixture of PBDEs, composed primarily of BDE-47, -99, and -100 (Crimmins et al. 2012). These congeners are found ubiquitously in North American freshwater, at concentrations typically more than 10 times higher than European values (Crimmins et al. 2012).
Prior to a voluntary U.S. industry phase out in 2004, 95% of worldwide pentaBDE was produced in North America (Crimmins et al. 2012

Conclusions and Implications
Current emissions of PAHs and PBDEs to the environment near urban and industrial centers continue to supply PBTs to Lake Superior. Although long-range atmospheric transport is a major pathway for PAH and PBDE distribution across Lake Superior, point sources along the lake perimeter significantly impact local and regional concentrations. Fluxes for both currently-emitted PBTs were generally into the lake near industrial and urban sources, whereas the open lake sites appeared to volatilize PBTs back into the atmosphere at much lower rates. Enhancing spatial and temporal coverage of the Great Lakes region will provide meaningful trends in PAH and PBDE concentrations and fluxes as use and regulations change. We anticipate PAH concentrations to remain constant or even increase with time, but PBDE concentrations may decrease as use is phased out. Changes in atmospheric concentrations could lead to an equilibrium reversal and the volatilization of PBDEs out of Lake Superior into the atmosphere. Such trends can be monitored effectively and affordably with polyethylene passive samplers.

Introduction
Organochlorine pesticides (OCPs) were produced and applied extensively in the United States from the 1950s to the 1970s (Shen et al. 2005). Primary application during that time was for agriculture, however pesticide use continued through the 1980s as termite control in residential areas (Shen et al. 2005). Thus, OCPs are found in rural and urban areas alike. Two insecticides which were widely used in North The largest agricultural regions in North America have historically been located to the south/southeast of the Great Lakes. OCPs are transported to Lake Superior by long-range atmospheric transport and deposited from the air into the surface water by wet and dry deposition, as well as by diffusive chemical exchange (Hillery et al. 1998;Swackhamer et al. 1999;Hafner and Hites 2003). Passive transfer of molecules, such as PBTs across the air-water interface is driven by the concentration gradient of an analyte, such that the additions of OCPs to surface water can reverse the gradient and result in volatilization (Hafner and Hites 2003). OCP volatilization is also strongly temperature-dependent, creating seasonal flux cycles with greatest release to the atmosphere at the warmest time of year (Cortes et al. 1998;Buehler et al. 2001).
OCP concentrations in atmosphere, water, and biota have been decreasing since regulatory controls were put into effect (Sun et al. 2006;Gouin et al. 2007;Gewurtz et al. 2008). Lake Superior's large surface area, long retention time, and colder water temps have allowed accumulation of these persistent pollutants over the past several decades (Buehler et al. 2004;Gewurtz et al. 2008). Reduced emissions from primary sources to the atmosphere have resulted in steady state and even net volatilization of "legacy" OCPs from the surface water to the overlying air (Hillery et al. 1998). Continued losses from both the air and water may lead to the virtual elimination of most OCPs from the Lake Superior environment by the mid-21 st century (Cortes et al. 1998). Pesticides in current or recent use, such as endosulfan I and its metabolites, may take longer to purge from Lake Superior.

Passive samplers have been effectively used to monitor OCP concentrations in
Great Lakes air (Gouin et al. 2007). Polyethylene passive samplers (PEs) rely on diffusion to measure time-integrated truly gas-and dissolved-phase analytes (Morgan and Lohmann 2008), excluding confounding influences from particulates, precipitation, and colloids. This fraction is important to the cycling of persistent pollutants because it is available for bio-uptake and free to passively exchange across the air-water boundary. PEs are an inexpensive alternative to active high-volume air sampling that can expand spatial and temporal monitoring of Lake Superior.
Additionally, PEs can be deployed simultaneously in air and water to determine the magnitude of OCP gradients across the air-water interface at each deployment site.
This study utilized PEs to measure gaseous and dissolved concentrations of 24 OCPs representing both legacy and recent applications. The goals of this monitoring were to (1) determine overall concentrations and distribution patterns across the lake and (2) establish the gradient and magnitude of OCP gas exchange across the air-water interface between mid-and late-summer.

Methods and Materials
PEs were prepared and deployed as discussed elsewhere (see Chapter 1).
OCPs were analyzed on a Waters Quattro micro GS Micromass MS-MS and quantified using Waters QuanLynx V4.1 software as detailed elsewhere (Zhang et al. 2012). Samples were held at 100 o C for 1 minute, ramped up to 220 o C at 5 o C/min and held for 10 minutes, then ramped up to 280 o C at 4 o C/min and held for 5 minutes.
Analysis was conducted in splitless injection mode using a 30m x 0.250mm i.d. (film thickness 0.25µm) DB-5MS column.
Quality assurance and control is discussed elsewhere (see Chapter 1). Average surrogate recoveries for were ca. 50% for 13 C6 HCB and ca. 170% for 13 C12DDT.
A total of 24 OCPs were included in quantification (see Supporting Information). Concentrations, flux gradients, and flux rates were determined according to equations presented elsewhere (see Chapter 1).

Geographic Information System Calculations (GIS)
The proximity of sampling sites to EPA Areas of Concern (AOCs) and National Priorities List Superfund sites (NPLs) was calculated using ArcMap 10.1 (see Supporting Information). Land use data for the United States was obtained from the Multi-Resolution Land Characteristics Consortium National Land Cover Database.
The most recent data available from 2006 was used to determine the amount of agricultural land within 10 km of sampling sites (see Supporting Information).

Hexachlorocyclohexane (HCH)
Technical HCH is a mixture of 5 HCH isomers used as broad-spectrum insecticides in the United States and Canada from the 1940s to 1970s (Buehler and Hites 2002). It consisted of 60% α-HCH, resulting in the release of ca. 6.5 Mt of the isomer between 1948 and 1997 (Shen et al. 2004). α-HCH was restricted in the United States and Canada in 1978 and has since been banned worldwide, but is still emitted into the environment during the production of lindane and other pesticides (Shen et al. 2005). Extensive use and long atmospheric residence times has led to ubiquitous α-HCH distributions globally (Hafner and Hites 2003). Relatively uniform distribution across Canada and the United States is expected for a persistent compound with no current use (Shen et al. 2004 and a virtual elimination in the atmosphere by ca. 2040. Since production was discontinued, atmospheric concentrations of α-HCH are no longer determined by primary sources, but are now mostly a result of evaporation from terrestrial and aquatic surfaces (Shen et al. 2004). Our data suggests that α-HCH is revolatilizing from Lake Superior, causing atmospheric concentrations over open lake waters to be slightly elevated relative to other lake regions.
The remainder of technical HCH was composed of 1-13% β-HCH and 10-15% γ-HCH, or lindane. These two isomers were not detected regularly across Lake Superior in this study. Although once extensively used in Canadian agriculture (Ma et al. 2004), lindane application was abandoned in Canada in 2004 (Sun et al. 2006).
Even before the ban, atmospheric γ-HCH had been declining since 1991 (Sun et al. 2006). γ-HCH has a slightly lower air-water partitioning coefficient than α-HCH and is therefore scavenged more efficiently (Shen et al. 2004) γ-HCH also reacts more readily in the atmosphere and, thus, is not transported as far as α-HCH (Shen et al. 2004). Since its disuse in 2004, atmospheric concentrations of γ-HCH may have declined by half according to an estimated half-life of ca. 8 years. Likewise, β-HCH has relatively low volatility and high stability resulting in distributions restricted to areas near point sources (Sun et al. 2006). β-HCH has the greatest physical and metabolic stability among all of the technical HCH isomers due to its relatively planar structure (Sun et al. 2006). These physicochemical properties may explain why β-HCH is not commonly detected in the atmosphere over Lake Superior. γ-HCH was detected at Ontonagon from August-October, while both γ-HCH and β-HCH were present at Ashland during the same period. This presence may indicate the historic use of HCH in those areas. According to 2006 National Land Cover Database maps, the Ontonagon site was within 10 km of ca. 180,000 acres of agricultural land, the greatest concentration within the same distance from all U.S. sites included in this study. Ashland was within 10km of ca. 73,000 acres of agricultural land (see Supporting Information).

Hexachlorobenzene (HCB)
HCB dominated in Lake Superior air, representing >80% of the OCPs measured at every site. HCB is a fungicide used in many applications in the past, but with no current commercial uses as an end product in North America (Buehler et al. 2004;Shen et al. 2005). Production was banned in the United States in 1965, but it is still formed as a byproduct in several manufacturing processes and is a known impurity in many pesticides currently used in the Great Lakes basin (Sun et al. 2006;Buehler et al. 2004). HCB has a very long atmospheric residence time, relatively high vapor pressure, low water solubility, and low sensitivity to hydroxyl radical attack, Though not significantly, concentrations were generally greater at Canadian shore and open-lake sites.

Endosulfan
Endosulfan I is a broad spectrum insecticide used on fruits, vegetables, cotton, tobacco, and trees, as well in the preservation of wood (Buehler et al. 2004;Shen et al. 2005). Use began in the 1950s and continued until recently (EPA 2010). Although now included in the Stockholm Convention and designated by the EPA for use to be terminated, it is one of the most abundant and ubiquitous OCPs in the continental atmosphere (Shen et al. 2005) and was widely used in the states surrounding Lake Superior (Hafner and Hites 2003). Application in Michigan, Wisconsin, and Minnesota resulted in the short-range atmospheric transport dominating inputs to the Great Lakes (Hafner and Hites 2003). Technical endosulfan is composed of two isomers, alpha and beta (Buehler et al. 2004). Previous studies have found alpha, endosulfan I, to be dominant in North American atmosphere (range 3.1-685 pg/m 3 ) (Shen et al. 2005). This pattern was consistent with the Lake Superior atmosphere, where concentrations were among the lowest across the continent (Shen et al. 2005 Buehler et al. (2004) was estimated when endosulfan was still in use; the decrease may be steeper now that emissions have been reduced, resulting in even faster elimination from the Great Lakes atmosphere. Endosulfan II was not significantly present at any of the Lake Superior sites included in this study. Endosulfan sulfate, an environmental breakdown product of endosulfan, was only detected in the atmosphere at a few sites.

Heptachlor epoxide
Heptachlor epoxide is the product of heptachlor degradation. Like heptachlor, it is persistent, bioaccumulative, and toxic (Bidleman et al. 1998), however, it is not currently included in the Stockholm Convention of POPs. Heptachlor is a chlorinated cyclodiene that was used for many years as an insecticide for agriculture, lawns, and gardens (Bidleman et al. 1998). Additionally, it was used as a termiticide, but most other applications were canceled in 1988 (Bidleman et al. 1998). Heptachlor is transformed into heptachlor epoxide by three main reactions: photolysis in the atmosphere, hydrolysis in the water, and epoxidation in soils and biota (Bidleman et al. 1998). Heptachlor epoxide formed by photolysis is racemic, whereas epoxidation from biological metabolism in agricultural soils produces nonracemic fractions enriched in the (+) enantiomer (Bidleman et al. 1998). Analysis of the enantiomeric ratio indicates that (+) enantiomer-enriched heptachlor epoxide concentrations measured in the Great Lakes region are mainly volatilized from soil rather than the product of photolysis (Shen et al. 2005). Secondary volatilization from soils suggests an aged source and we would expect a fairly uniform distribution.
Heptachlor epoxide was detected in the atmosphere at all sites across Lake Superior at an average of 0.62 pg/m 3 (range 0.29-1.1 pg/m 3 ). Concentrations were low and fairly uniform across the lake, in-line with its volatilization from soils. Greatest concentrations were measured at Thunder Bay/Welcome Isle and Sault Saint Marie (1.1 pg/m 3 and 0.96 pg/m 3 , respectively), two populated areas where heptachlor may have been more recently and extensively applied for termite control. Secondary sources are also suggested by inconsistent detection of heptachlor, which is more volatile than heptachlor epoxide and is expected to dissipate more quickly by evaporation (Bidleman et al. 1998). The parent compound was present at fewer than half of the sites, mainly elevated near populated and industrial areas, especially Thunder Bay/Welcome Isle (1.2 pg/m 3 ).

Chlordane
Technical chlordane (trans-chlordane, cis-chlordane, and trans-nonachlor) was used as an insecticide, herbicide, and termiticide between 1947 and 1988 at which time it was deregistered for all uses in the United States, and 1990 when it was banned for all uses in Canada (Hafner and Hites 2003;Shen et al. 2005;IADN 2008

Dichlorodiphenyltrichloroethane (DDT)
Technical DDT (65-80% p,p'-DDT, 15-21% o,p'-DDT,<4% p,p'-DDD) was used extensively in the 1940s and 1950s in urban aerial sprays to control mosquitoes (Sun et al. 2006) and widely used on a variety of agricultural crops in the 1960s (Shen et al. 2005). It was deregistered in the United States in 1972 and in Canada in 1973 (IADN 2008). DDTs are highly persistent and residues in urban locations may still be sources of these compounds to the local atmosphere (Sun et al. 2006). Because DDTs are less volatile than most OCPs, their atmospheric transport is limited and concentrations tend to be greater near urban sources (Shen et al. 2005;Sun et al. 2006). Thus, unlike other long-banned OCPs, DDTs do not exhibit a uniform distribution. DDT was the first pesticide to be banned and likely has the fewest current sources in the environment (Cortes et al. 1998). Vapor phase concentrations measured over the past few decades have demonstrated a significant decreasing trend at most sites across the Great Lakes (Cortes et al. 1998;IADN 2008). Technical DDT compounds measured in this study were present across Lake Superior at very low concentrations (<0.4 pg/m 3 ). p,p'-DDT was often below detection and p,p'-DDD+o,p'-DDT was detected at low concentrations at every site (0.02-0.37 pg/m 3 ).
These low concentrations follow a long-term decreasing trends in vapor-phase concentrations, consistent with a reported half-life of 5-17 years (Buehler et al. 2004;IADN 2008). p,p'-DDT had the earliest virtual elimination date of the OCPs measured (ca. 2010) (Cortes et al. 1998). However, p,p'-DDT dechlorinates in the environment to form the metabolites p,p'-DDE and p,p'-DDD (Lohmann et al. 2009;IADN 2008). p,p'-DDD is removed from the atmosphere at the same rate as p,p'-DDT (Cortes et al. 1998), but p,p'-DDE has higher vapor pressure and higher Henry's law constant, preferentially partitioning into the gas phase, extending its presence in the atmosphere (Sun et al. 2006 Harbor, so we would expect dieldrin to continue to be detected in the atmosphere above Lake Superior as well as in the water. Aldrin and its other metabolite, endrin, were not detected in the water during this study.

Temporal Air Distributions
It has been well-established that most atmospheric concentrations of semivolatile compounds vary seasonally with temperature, resulting in peak concentrations during summer months (Cortes et al. 1998;Ackerman et al. 2008;Hillery et al. 1998).
Previous studies found atmospheric α-HCH was 4-6 times greater in the summer compared to the late fall and early spring (Perlinger et al. 2005). In this April-October study, atmospheric OCP concentrations were generally consistent, perhaps exhibiting slight increases during August-October when average air and water temperatures were highest.

Overview
Marquette and Sturgeon Bay were not measured from August-October, so are excluded from spatial averages. Only 8 OCPs were measured above the detection limit at all sites from June-October: α-HCH, HCB, heptachlor epoxide, trans-and cischlordane, trans-nonachlor, p,p'-DDE, and combined p,p'-DDD+o,p'-DDT ( Figure   3.3). Endosulfan I was not consistently measured in the water. Dissolved OCP concentrations in Lake Superior surface water were dominated by α-HCH (average 250 pg/L), followed by HCB (average 16 pg/L). Heptachlor epoxide was present at fairly uniform concentrations across the lake (average 3.6 pg/L). Although not detected at every site, dieldrin was also significantly present at most coastal sites (average 24 pg/L). Atmospheric transport has historically been the input of OCPs to Lake Superior (Baker and Eisenreich 1990), rather than from inflow of contaminated June-August and August-October. Because α-HCH is no longer directly emitted into the atmosphere, stores of the compound in the surface water of Lake Superior are susceptible to influences from seasonal temperature fluctuations. A strong response to seasonal variation has been well established (Shen et al. 2004;Perlinger et al. 2005).
Lake Superior's large surface area (82,100 km 2 ) and long water residence time (191 years), coupled with α-HCH's high volatility may result in greater volatilization from the lake surface to the overlying atmosphere with increasing ambient temperatures (Gewurtz et al. 2008).

Hexachlorobenzene (HCB)
HCB was also present at all monitored sites, and exhibited similar distribution patterns as α-HCH. Coastal concentrations were fairly uniform, most ranging from 5.4 to 14 pg/L. Again, the Duluth water had a slightly greater average (18 pg/L), approaching the elevated concentrations of the open lake sites (average 23 pg/L).
Ashland was an outlier where HCB was 37 pg/L. In addition to its application to agricultural seeds as an antifungal agent, HCB was also emitted in the waste streams of wood-preserving plants and the incineration of municipal waste (IADN 2008

Dichlorodiphenyltrichloroethane (DDT)
p,p'-DDT was not measured above the detection limit at every site. Where (0.43 ng/g) were also greater than those of DDT (0.11 ng/g) (Gewurtz et al. 2008).
Although at low concentrations, DDTs are persisting in Lake Superior and, as POPs, are known to have the potential to bioaccumulate and cause toxic effects on non-target organisms (Gouin et al. 2007). In a study of pesticides in Western U.S.
National Park fish conducted from 2003-2005, p,p'-DDE was one of the most concentrated semi-volatile contaminants measured in >75% of the fish (Ackerman et al. 2008). DDTs as a whole were 2-5 times greater in western U.S. fish than in oceancaught salmon. In Lake Superior DDE reached up to 81+-19 ng/g in bloater fish (Kucklick and Baker 1998).

Dieldrin
Dieldrin was regularly detected in the water, in stark contrast to the air, where dieldrin was not consistently detected. Dieldrin represented 10-20% of total dissolved OCPs measured in Lake Superior water. Excluding Ontonagon, Duluth, and the open lake sites where the analyte was not present above the detection limit, average dieldrin Concentrations in Lake Superior bloaters were similar to DDE concentrations (Kucklick and Baker 1998).

Endosulfan
Endosulfan I was not consistently detected in Lake Superior water, only averaging 0.63 pg/L. Endosulfan sulfate, an endosulfan metabolite, was detected at a few sites at high concentrations (44-360 pg/L), possibly indicating a tendency for endosulfan to degrade in Lake Superior water, however measurements were very inconsistent and do not illustrate any definitive trends. Endosulfan sulfate was among the most concentrated and frequently detected contaminants measured in western U.S.
National Park fish (Ackerman et al. 2008) and may be a threat to Lake Superior fish as endosulfan I continues to be deposited and degrade into the metabolite.

OCP Air-Water Gradient
Equilibrium concentration ratios between air and water for α-HCH were <1 at most sites, indicative of net volatilization. However, ratios were only significantly different from equilibrium (>3.9 or <0.14) at ca. 1/3 of the sites (Figure 3.4A), suggesting atmospheric deposition may still play a role in air-water exchange, or that α-HCH is reaching equilibrium. α-HCH deposition into Lake Superior has been declining for several years (Cortes et al. 1998;Hillery et al. 1998;Buehler and Hites 2002), allowing for an equilibrium shift across the air-water boundary. Over the past decade, α-HCH volatilization from Lake Superior surface water has been observed (Shen et al. 2004;Gouin et al. 2007). As α-HCH concentrations in the air continue to decline, we expect net volatilization to become significant across the entire lake. HCB was also volatilizing from half of the lake sites, but only significantly <1 at 2 sites continue to be added to Lake Superior. Due to their persistence in the environment, current deposition may be part of a long-term cycling of DDTs across the air-water boundary as indicated by previous studies (Hillery et al. 1998).

OCP Air-Water Exchange Fluxes
The OCPs with the greatest concentrations in either air (deposition) or water (volatilization) displayed the greatest flux rates across the air-water boundary in Lake Superior. Average net volatilization of α-HCH across the entire lake from June-October was 0.70 ng/m 2 /day (range below detection to 3.5 ng/m 2 /day), with nearly two-thirds of the sites exhibited fluxes not significantly different from equilibrium.
Ontonagon was an exception, where α-HCH appeared to be deposited at a rate of -0.82 ng/m 2 /day during August-October (Figure 3.4A).
Gaseous HCB fluxes were much greater than those of gaseous α-HCH.
Average flux rates across the lake from June-October for HCB were only significantly indicative of volatilization at Ashland (110 ng/m 2 /day) and Ontonagon (25 ng/m 2 /day) ( Figure 3.4B). Open lake sites and Eagle Harbor were either at, or near, equilibrium.
In fact, flux rates from 2002-2003 found HCB to still be depositing at a rate of -14 to -2.2 ng/m 2 /day while α-HCH was ca. equilibrium (Perlinger et al. 2005).
Endosulfan I was significantly deposited at nearly every site included in this study, at an average of 0.013 ng/m 2 /day ( Where present, dieldrin was volatilizing from the lake at an average rate of 0.46 ng/m 2 /day from June-October. Fluxes out of the lake appeared to double between mid to late summer (Figure 3.4D). At these rates, the lake may be serving as an important secondary source of dieldrin to the atmosphere. OCPs with recent emissions, such as endosulfan I, are still undergoing atmospheric transportation to the lake, resulting in strong net deposition across the entire lake surface. Continued monitoring is required to determine the long-term effects of regulation and fate of these compounds in the Lake Superior region.

Conclusions and Implications
Polyethylene passive samplers make it possible to easily and affordably monitor continuing OCP trends at a high resolution, distinguishing between background concentrations likely transported over long distances and local influences from populated areas. PEs should be deployed year-round for the next several years in order to fully establish seasonal and annual cycles in addition to long-term trends. Calculations are based upon site averages from the second (June-August) and third (August-October) deployments a Excludes Marquette and Sturgeon Bay because not present during both June-August and August-October 2011 b Analytes with medians below detection were excluded

Introduction
Polychlorinated biphenyls (PCBs) were used extensively in United States industry from the 1930s through the 1970s (Buehler and Hites 2002). They were one of the original persistent organic pollutants added to the Stockholm Convention and are known to bioaccumulate and cause adverse, irreversible effects in fish, birds, and mammals in the Great Lakes region (Suchash et al. 1999). Prenatal exposure to PCBs in the Great Lakes region has resulted in lower IQ and provided evidence that PCBs are neurobehavioral toxicants (Stewart et al. 2003;Stewart et al. 2008). Although production was banned in 1979 (Honrath et al. 1997), significant amounts of PCBs are still present in closed systems as dielectric fluid in electrical transformers and capacitors, in addition to PCB waste in landfills (Suchash et al. 1999). The majority of known PCB releases to the land from 1990-1999 were from transformer or capacitor liquid spills contributing to a total release of 278 pounds to the air and 2,621,169 pounds to the land (U.S. PCB Emissions Inventory 2012).
PCBs are largely urban pollutants (Buehler and Hites 2002), however their resistance to degradation has allowed them to develop a ubiquitous presence in the environment (Suchash et al. 1999). Most sources are located to the south of the Great Lakes, resulting in greater atmospheric concentrations over Lake Superior when winds transport PCBs from the south (Suchash et al. 1999). PCBs have been detected in Lake Superior air, water, and sediment for several decades, but concentrations are generally lower than the other Great Lakes, likely due to distance from major urban areas (Gewurtz et al. 2008).
Atmospheric deposition is estimated to provide 90% of PCBs in Lake Superior . Slow degradation and benthic recycling Jeremiason et al. 1998) allows PCBs to remain in the water column after deposition. A reduction in PCB emissions from primary sources to the atmosphere results in a measured revolatilization from the lake and peak concentrations in the air during the warmer months (Buehler et al. 2011). This flux has resulted in a buffered level of PCBs in the atmosphere, whereas surface water and biota concentrations have decreased Hillery et al. 1997).
This study aims to increase the sampling area of Lake Superior in order to understand the dynamics of PCB concentrations detected in the lake. Long-range atmospheric transport from industrial areas in the south are expected to result in relatively even distribution of PCBs across Lake Superior. IADN conducts long-term, year-round sampling, generating an invaluable dataset of PBT distributions on the Great Lakes, but sampling sites, such as Eagle Harbor on Lake Superior, were selected to represent remote, background levels. Omission of urban sampling overlooks local sources along the coast of Lake Superior which have already demonstrated elevated PCB concentrations in offshore sediments (Eisenreich and Hollod 1979;Gewurtz et al. 2008). Although limited to one season, this study was conducted to contribute to monitoring of PCBs as they cycle and decline in the environment, as well as increase the spatial resolution of PCB measurements in order to more accurately represent the distribution and fluxes of these persistent pollutants across Lake Superior. We also aim to demonstrate the efficacy and efficiency of polyethylene passive sampling as a useful tool for long-term, time-integrated PCB measurements.

Methods and Materials
PEs were prepared and deployed as discussed elsewhere (see Chapter 1).

OCPs were analyzed on a Waters Quattro micro GS
A total of 18 PCBs were included in quantification (see Supporting Information). Concentrations, flux gradients, and flux rates were determined according to equations presented elsewhere (see Chapter 1). Proximity of sampling sites to potential source areas was determined using ArcMap 10.1 (see Supporting Information).

PCB Air Distribution
PCBs were detected in the atmosphere at nearly every site during the sampling season, however distributions were not spatially and temporally uniform. Sault Saint Marie and Marquette had the highest Σ18PCB concentrations from June-August, >15x higher than the other sites (Table 4.1). These two sites also had the greatest number of different PCB congeners present (Figure 4.1A), suggesting that Sault Saint Marie and Marquette are current sources of PCBs, probably due to their historical use at both locations. In general, PCB concentrations were higher and more diverse near populated or industrialized areas, as expected for anthropogenic products with no known natural emissions (Hillery et al. 1997;Honrath et al. 1997;Hafner and Hites 2003). Despite associations with larger populations, PCB concentrations were relatively low at the Duluth and Thunder Bay/Welcome Isle stations. As discussed previously, the Duluth site is located northeast of the downtown, and may not receive Reduced anthropogenic emissions to the environment may explain the current PCB trends in the air above Lake Superior. After production was banned in 1979, PCBs over Lake Superior decreased to North American background levels (ca. 1 ng/m 3 ) and remained fairly stable for several years (Baker and Eisenreich 1990;Hillery et al. 1997;Jeremiason et al. 1998). A number of studies from the past 20 years have suggested that gaseous PCB concentrations were maintained in the Great Lakes' atmosphere due to replenishment from the lakes themselves (Baker and Eisenreich 1990;Cortes et al. 1998;Buehler et al. 2004). It was hypothesized that the lakes act as buffers, responding to the reduced atmospheric loadings and releasing PCBs back to the air. This volatilization in turn would lead to lower PCB concentrations in the surface water (Baker and Eisenreich 1990;Jeremiason et al. 1994). Air-water exchange is dominated by the more volatile tri-and tetrachlorobiphenyls. Their vulnerability to OH radical degradation (Gevao et al. 1998) and subsequent continued volatilization may cause less-chlorinated PCBs to be preferentially removed from the Lake Superior atmosphere over time. Although triand tetrachlorobiphenyls were historically the dominant congeners in the Great Lakes' atmosphere (Baker and Eisenreich 1990;Gevao et al. 1998), in their absence, pentaand hexachlorobiphenyls may appear to dominate. PCB levels in the gas phase have continued to decline significantly (Gewurtz et al. 2008).

Current atmospheric PCB levels indicate a decrease in gaseous PCBs in Lake
Superior air over the past decade. Previous IADN studies reported ΣPCB at Eagle Harbor ca. 63-95 pg/m 3 from 1990-2003 (Hillery et al. 1997;Buehler et al. 2001;Sun et al. 2007). We found concentrations from June-October to be lower by an order of magnitude or below the detection limit, exceeding the 18+-7.1 year estimated half-life for PCBs at Eagle Harbor (Buehler et al. 2004). A number of factors in addition to reduced emissions could have contributed to the magnitude of the apparent decrease.
Previous studies all reported ΣPCB for more than 18 congeners, possibly adding significant fractions not included in this study. Also, IADN samples were collected using a high-volume active air sampler, which may produce overestimates due to particle desorption on the filter (Perlinger et al. 2005). Additionally, air sampling rates in this study were calculated to be 8-120 m 3 /day. Overestimates for the air sampling rate in our study may have resulted in underestimates for PCB concentrations. Conversely, sampling for this study was conducted during the warmest time of year and should have resulted in overestimates compared to the annual average reported by IADN studies. Previous studies indicated enhanced volatilization during the summer resulting in annual maximum vapor-phase concentrations in July/August .

PCB Water Distribution
PCBs were detected at every site from June-August ( PCB concentrations in the Lake Superior water column have been declining for the past few decades (Baker and Eisenreich 1990;Buehler and Hites 2002), removed primarily by volatilization (Jeremiason et al. 1998). Sedimentation accounted for 4900 kg of PCB removal from the surface water between 1980 and 1992, however volatilization to the atmosphere removed 26,500 kg of PCBs over the same time period . In 1980 dissolved concentrations in Lake Superior were 2.4 ng/L. This level was reduced to 0.18 ng/L by 1992 at a first-order decay of 0.2/year . Dissolved PCB concentrations from this study indicate an elevated rate of removal, perhaps in response to decreased loadings.
Overall half-lives for PCBs in both the air and water were estimated to be 5-9 years (Hillery et al. 1998). Applying this rate to 1992 levels yields results similar to urban and industrialized coastal sites measured in this study, but are much higher than rural areas. These results suggest that developed areas continue to have an impact on local PCB concentrations.
Since atmospheric deposition is considered to be the main input mechanism, accounting for 85-90% of PCB loading into Lake Superior (Eisenreich and Hollod 1979;Honrath et al. 1997), decreased anthropogenic emissions have resulted in reduced levels detected in Lake Superior biota (Hillery et al. 1997;Jeremiason et al. 1998). However, since PCBs bioaccumulate, they may be maintained in higher trophic organisms and remain a potential health hazard (Wong et al. 2004). Lake Superior trout exhibited ΣPCB concentrations ca. 5440 ng/g lipid (Wong et al. 2004).
Penta-, hexa-, and heptachlorobiphenyls tend to be dominant in biota, similar to congener patterns observed in the water column (Kucklick and Baker 1998;Wong et al. 2004). Some PCB removal may occur through biotransformation and sinking of dead organic matter (Kucklick and Baker 1998;Wong et al. 2004), but it has been shown that Lake Superior experiences efficient carbon recycling, whereby the pelagic food chain and benthic food web maintain PCBs in the water column (Jeremiason et al. 1998). Despite the hydrophobic nature of PCBs and the sinking of >50% of total PCB stores in Lake Superior on settling particles, only 2-5% of PCBs accumulate in bottom sediments Jeremiason et al. 1998). Current surface water concentrations indicate that PCB levels in biota should continue to decrease, but given that the truly dissolved fraction only represents half of the PCBs present in the water column (Baker and Eisenreich 1990), PCBs are expected to continue to be detected in Lake Superior biota, especially in higher trophic level organisms.

PCB Air-Water Gradient
For the PCB congeners present, concentration ratios between the air and water were predominantly less than one at all sites from June-October. Such low ratios indicate that volatilization was dominating air-water exchange across Lake Superior.
Two coastal sites exhibited net deposition for tetra-and pentachlorobiphenyl from June-August: Marquette and Thunder Bay. Marquette is an industrial town with an AOC and remediated NPL, and may be acting as a source of PCBs to the atmosphere, enhancing deposition in the nearby water compared to other Lake Superior sites.
Likewise, Thunder Bay is the largest city on Lake Superior and may still be a source of PCBs, even though atmospheric concentrations were much lower than at Marquette. Duluth did not exhibit the same deposition, likely because the air site was located northeast of the city emissions, whereas water samples were collected a few kilometers from the mouth of the St. Louis River. Foster Island was also dominated by PCB deposition. Although overall atmospheric and water concentrations were very low at Foster Island, it may be receiving small inputs from the nearby Superfund site at Peninsula Harbor.
Open lake sites had ratios indicating strong volatilization of tri-, tetra-, and pentachlorobiphenyls, and equilibrium for heavier congeners. This pattern is expected for regions with no direct local PCB inputs, where fluxes are controlled by environmental factors, such as ambient temperature and wind speed. Particle scavenging efficiently removed PCBs from the atmosphere, creating an urban to rural gradient in PCB distribution (Buehler et al. 2001;Offenberg and Baker 2002), and possibly limiting the amount and weight of congener transported to the open lake.
Additionally, temperature differences between the land and water in the summer cause enhanced offshore air flow, which, in turn, creates a stable atmosphere (Honrath et al. 1997). Reduced atmospheric mixing hinders the interaction of land-based pollutants with the water surface, limiting transportation to within ca. 30 km from shore. The thinner layer free to exchange PCBs across the air-water boundary quickly equilibrates with the lake surface, supporting the flux gradients observed in the open lake.
Increased temperatures from August-October corresponded with an enhanced volatilization signal for most congeners across the lake. Temperature changes were greatest at the open lake sites (average 6.6 o C), likely causing volatilization strong enough to overwhelm deposition of higher molecular weight PCBs. Congeners ranging from penta-to heptachlorobiphenyls had ratios suggesting strong volatilization from open lake surface water. A similar switch from net deposition to net volatilization was observed for Foster Island at a more modest air temperature increase (2.8 o C). Changes in gradients may also be caused by reduced atmospheric sources.
The Marquette water sampler from August-October was lost, but Thunder Bay/Welcome Isle exhibited ratios similar to June-August. Ontonagon changed the most, exhibiting strong net deposition of tri-and tetrachlorobiphenyls in response to the elevated atmospheric concentrations of those congeners during the same time period. It is unclear from where the light PCBs originated. Similar to most sites across Lake Superior, heavier congeners present (penta-, hexa-, and heptachlorobiphenyl) were significantly volatilizing from the surface water at this site.
These results follow a trend of reported PCB volatilization from Lake Superior over the past 20 years, the greatest observed fluxes occurring in July-September (Baker and Eisenreich 1990;Hornbuckle et al. 1994). The current study was limited to April-October, restricting its scope to the warmest times of the year. Therefore, it cannot be determined whether these flux gradients are representative of annual airwater exchange, or if net deposition dominates PCB flux during the colder months.
We do see a shift in the behavior of PCBs by degree of chlorination compared to previous measurements. Two decades ago, the less-chlorinated congeners ( 5 chlorines) exhibited volatilization fluxes throughout the year, representing 90% of PCB volatilization, whereas highly chlorinated congeners had depositional fluxes . Our flux gradients suggest that volatilization is dominated by hexachlorobiphenyls in the general absence of tri-and tetrachlorobiphenyls.  (Figure 4.3C). Tri-and tetrachlorobiphenyls were no longer volatilizing from the Duluth site in August-October because they were not detected in the water. Volatilization may have removed sufficient amounts of the lower molecular weight congeners from the water to the atmosphere to reduce water PCB concentrations below the limit of detection.

PCB Air-Water Exchange
We did not observe a corresponding increase in atmospheric PCB concentrations for Duluth, likely due to the horizontal distance between the air and water sampling sites.
Rural coastal sites exhibited lower PCB flux rates, supporting previously mentioned urban-rural gradients in PCB distributions (Figure 4.3D, 4.3E). An Net deposition of tetra-and pentachlorobiphenyls dominated flux rates at Marquette from June-August (-12 ng/m 2 /day). There is no data for this site from August-October. Ontonagon showed net deposition of tri-and tetrachlorobiphenyls from August-October (-14 ng/m 2 /day).
Average flux across the lake from June-October was ca. 4.6 ng/m 2 /day (range -12 ng/m 2 /day to +24 ng/m 2 /day). The range is lower than the annual average reported for 1978 to 1992, 63 ng/m 2 /day . Since sampling for this study was limited to the warm summer months, it was expected that average volatilization rates would be greater than for annual averages where deposition may be enhanced in cooler winter months. Overall, water-air exchange appears to be decreasing, consistent with decreasing rates of volatilization observed in the 1990s (Buehler and Hites 2002).

Conclusions and Implications
PCB concentrations measured in the air and water across Lake Superior exhibited a clear association with urban and industrial areas, consistent with past findings in the Great Lakes region. However, current concentrations appear to be dominated by penta-and hexachlorobiphenyls, whereas past measurements found triand tetrachlorobiphenyls to make the greatest contributions to PCB levels. Eagle Harbor consistently had the lowest PCB concentrations. Atmospheric concentrations at open lake sites were similar to rural coastal areas from June-August, however dropped below the detection limit in August-October. This decrease was likely because net volatilization was stronger in mid-summer, but was not observed in latesummer, curbing the supply to the air. Likewise, PCB concentrations in the open water from June-August were higher than at rural coastal sites, but similar from August-October, perhaps as a result of a greater loss to the atmosphere from the open lake than from coastal waters.
Tandem air-water sampling of 14 sites across the lake demonstrated the efficacy of PEs to enhance the understanding of PCB dynamics in the Lake Superior air and water. They also provided an affordable and easy means of determining simultaneous, time-integrated equilibrium concentrations between the gaseous and dissolved phases and, thus, air-water exchange of PCBs. PEs should be used in concert with active high-volume sampling to enhance the spatial resolution of ongoing PCB monitoring in the Great Lakes region.

Sampling proximity to EPA Areas of Concern, National Priority List locations, and agricultural activity
The U.S.-Canada Great Lakes Water Quality Agreement (Annex 1 of the 2012 Protocol) defines an AOC as "a geographic area designated by the Parties where significant impairment of beneficial uses has occurred as a result of human activities at the local level" (GLWQA 2012). There are 8 AOCs in the Lake Superior region (FIGURE SI 1). National Priorities List areas are EPA-designated "superfund" sites affected by hazardous substances, pollutants, or contaminants requiring long-term remedial action. There are 6 NPL sites along the U.S. coast of Lake Superior. Proximity of these impaired locations to PE deployment sites were determined using the ArcGIS (10.1) "Near" tool (TABLE SI 1). U.S. state and Canadian territory data was obtained from the USGS (geonames.usgs.gov/domestic/download_data.htm). All data was converted to the geographic coordinate system GCS_North_American_1983 and projected in the projected coordinate system NAD_1983_UTM_Zone_16N. The image map from NLCD was converted to a raster and clipped to a 10-kilometer buffer created around each sampling site. Agricultural use was selected in the attribute table and the number of pixels with that use were calculated using the "Zonal Statistics as Table" tool. Pixel counts were converted to area according to the 30x30meter pixel size.

Concentration calculations
Analyte responses from GC/MS analysis are converted to concentrations (ng/PE) by: CPEa = (Ra / Rs) * (Cs / slope of standard curve) where Ra is the analyte response, Rs is the surrogate response, and Cs is the concentration of surrogate added. This calculation corrects for losses in recovery. The percent of recovery is determined by %recovery = CPEs /Cs where CPEs (pg/PE) is the concentration of surrogate recovered in the PE after sample processing CPEs = (Rs / Ri) * (Ci / slope of standard curve)

Gas exchange between the atmosphere and water
Calculations of air-water exchange are based on a modified two-film resistance model. where CeqW is the PBT concentration in the PE when in equilibrium with the surrounding water and CeqA is the PBT concentration in the PE when in equilibrium with the surrounding air. The reciprocal of kol is the sum of the resistance to mass transfer in the air and water: 1/ kol = 1/ kw + 1/( ka * H') where kw is the air-side mass transfer coefficient, ka is the air-side mass transfer coefficient and H' is the Kaw, or, the dimensionless Henry's Law constant. Temperature-corrected Kaw values for PAHs were calculated following Ma et al. (2010). Temperature-corrected Kaw values for PBDEs were calculated by H/(R * T) Where H is the temperature-corrected Henry's Law constant from SPARC (https://archemcalc.com/sparc), R is the gas constant, and T is the sampling temperature. The rate of transfer is related to the molecular diffusivity, thus kw and ka for specific compounds can be predicted by kw = kw(CO2) * [Sc / Sc(CO2)] -1/2 ka = ka(H2O) * [Da / Da(H2O)] 0.67 where kw(CO2) = 0.45 * u10 1.64 ka(H2O) 0.2 * u10 + 0.3 u10 is the windspeed at ten meters above the water surface obtained from the NOAA National Data Buoy Center (http://www.ndbc.noaa.gov/). Sc is the Schmidt number calculated for each analyte estimated from Schwarzenbach et al. (2 nd Ed. 2003) by dividing the kinematic viscosity of water by the temperature-dependent molecular diffusivity of the compound in water (Dw, cm 2 /s) obtained from SPARC (October 2012, https://archemcalc.com/sparc). Da is the temperature-dependent diffusivity of individual compounds in air. Both Da and Da(H2O) were obtained from SPARC.

Sampling rates
Sampling rates were determined using performance reference compound equilibrium according to Booij and Smedes (2010): Derived sampling rates are reported in Table SI 2. Samplers deployed in the water at Canadian coastal sites were housed in steel "spider" carrier cages (Environmental Sampling Technologies, http://www.est-lab.com/spmd.php). Open lake water samplers deployed at Environment Canada buoys were housed in open-ended copper pipes. All other water samplers were attached to moorings using only a single stainless steel wire, possibly explaining the higher sampling rates.                           Yunker et al. (2002). Ant/Ant+Phn <1 indicates a petroleum source rather than combustion. The outlying ratio at 0.11 is from Ashland and is indicative of gasoline combustion. Flra/Flra+Pyr >0.5 indicates kerosene, grass, coal, and wood combustion, while a ratio <0.5 indicates gasoline, diesel, fuel oil, and crude oil combustion. BaA/BaA+Chry >0.35 indicates a combustion source, while a ratio of 0.2-0.35 could indicate either petroleum or combustion. Ratios <0.2 indicate a petroleum source. Lake Superior samples yield ratios that suggest a mixture of petroleum and diesel/oil combustion.               Komp and McLachlan (1997) f Morgan and Lohmann (2008) g Recognized average value (Lohmann 2012)