In Situ Investigation of Performance Reference Compound-Based Estimates of PCB Equilibrated Passive Sampler Concentrations and C_free in the Marine Water Column

Abstract

Low-Density polyethylene (PE) sheets are used as passive samplers for aquatic environmental monitoring to measure the freely dissolved concentration (C_free) of hydrophobic organic contaminants (HOCs). Freely dissolved HOCs in water will partition into the PE until a thermodynamic equilibrium is achieved; that is, the HOC’s activity in the passive sampler is the same as its activity in the surrounding environment. One way to evaluate the equilibrium status or estimate the uptake kinetics is by using performance reference compounds (PRCs). A fractional equilibrium (f_eq) can be determined for target HOCs, under the assumption that PRC desorption from the passive sampler occurs at the same rate as the unlabeled target HOCs. However, few investigations have evaluated how effectively and accurately PRCs estimate target contaminant C_free under in situ conditions. In this study, PE passive samplers were pre-loaded with six, ¹³C-labelled PCBs as PRCs, and deployed in New Bedford Harbor, MA, USA and were collected after 30, 56, 99, and 129-day deployments. Using this unique temporal sampling design, PRC results from each deployment were fit to a diffusion model to estimate the C_free of 27 PCB congeners and compare the results between the different deployment times. Smaller PCBs had variable concentrations over the four deployments while mid-molecular weight PCBs had consistent C_free measurements for all deployments (relative standard deviation < 20%). High molecular weight PCBs had the largest C_free estimates after 30 days, these estimates and their standard deviations decreased with longer deployment times. These findings suggest when targeting PCBs with more than six chlorines or contaminants with a log K_OW ≥ 6.5, a deployment time longer than 30 days may be prudent.

Introduction

Passive sampling use has increased for scientific and regulatory applications (Lydy et al. 2014; Booij et al. 2016) due to the range of data it can provide about targeted contaminants and how this information can be used by environmental managers to make informed decisions. Useful information gleaned from passive samplers about targeted contaminants includes the likelihood of bioaccumulation by aquatic organisms (Joyce et al. 2016); transfer between environmental phases (e.g., sediments to water to atmosphere (Khairy et al. 2015; Liu et al. 2016; McDonough et al. 2016); and freely dissolved concentrations (C_free) – a critical value for understanding exposure and bioavailability (Mayer et al. 2014). Low-density polyethylene (PE) is a commonly applied passive sampler due to its wide availability, low expense, structural simplicity, and relative ease of use (Adams et al. 2007; Lohmann et al. 2012). The most straightforward way to measure C_free is to allow the contaminants to reach a thermodynamic equilibrium with the passive sampler and surrounding water. C_free can then be calculated from a passive sampler relatively simply according to Equation 1. The C_free of a targeted contaminant at equilibrium between the sampling environment and polyethylene (PE) can be calculated from:

$$C_{free}= \frac{C_{PE}^{\infty }}{K_{PEW}}$$ [1]

where, K_PEW is the PE-water partition coefficient for a specific contaminant and C_PE^∞ is the contaminant’s concentration in the PE passive sampler at equilibrium. As dictated by Equation 1, accurate estimates of C_free are contingent on an accurate measurement of C_PE^∞ and K_PEW values. Often, during in situ deployments it is unclear if a thermodynamic equilibrium was attained, and for more hydrophobic compounds, it may take several months to years to attain equilibrium (Fernandez et al. 2012, Ghosh et al. 2014). In these instances, the concentration in the sampler can be expressed as C_PE^t where the ‘t’ signifies the deployment duration and mass transfer calculations can be used to estimate C_free.

Various methods use performance reference compounds (PRCs) to model the mass transfer kinetics which allows measurement of C_PE^t to estimate C_PE^∞ or C_free directly (Booij, Sleiderink, and Smedes 1998; Booij and Smedes 2010; Fernandez, Harvey, and Gschwend 2009; Tcaciuc, Apell, and Gschwend 2015; Thompson, Hsieh, and Luthy 2015; Vrana and Schüürmann 2002; Shen and Reible 2019). Often, a PRC is an isotopically labeled direct analog of a target compound and the two compounds are assumed to have isotropic uptake and release kinetics. Using physiochemical characteristics of the PRCs and target compounds the corrections for target compounds without a direct isotopic analog PRC can be interpolated from the available PRC data. The PRCs are preloaded into the passive sampler prior to deployment and have similar partitioning and kinetic characteristics to target compounds but are also analytically distinguishable and not present in the sampled environment. The fraction of a PRC retained in the PE over the deployment (f_ret) allows for the calculation of site- and compound-specific sampling rates or estimation of the size of the water boundary layer (WBL) such that a fractional equilibrium (f_eq) can be calculated for several target contaminants without directly analogous PRCs based on log K_OW or molecular volume. PRCs are of particular value in field deployments of passive samplers (i.e., in situ) where equilibrium cannot be assumed or determined in contrast to laboratory, or ex situ deployments where conditions can be controlled to favor achieving equilibrium. Several studies have used PRCs in water column and sediment in situ deployments to estimate C_free (Burgess et al. 2015; Perron et al. 2013a, 2013b; Fernandez, Harvey, and Gschwend 2009; Fernandez et al. 2012; Estoppey et al. 2014; Tomaszewski and Luthy 2008; Apell and Gschwend 2016; Smedes 2007; Joyce et al. 2015; Apell et al. 2018; Sanders et al. 2018). In a previous study, we suggested that using a sampling rate approach (Booij and Smedes 2010) or a diffusion-based approach (Thompson, Hsieh, and Luthy 2015) to determine C_free often resulted in statistically indistinguishable results in a field deployment (Joyce and Burgess 2018). In this investigation, we used PRCs to populate a diffusion-based model to estimate f_eq and calculated C_PE^∞. This model is based on the governing equations:

$$\frac{\partial {\text{C}}_{\text{PE}}}{\partial \text{t}}={\text{D}}_{\text{PE}}\frac{{\partial }^2{\text{C}}_{\text{PE}}}{\partial {\text{x}}^2}\text{for}0<\text{x}<1$$ (2)

$$\frac{\partial {\text{C}}_{\text{WBL}}}{\partial \text{t}}=D_{\text{W}}\frac{{\partial }^2{\text{C}}_{\text{WBL}}}{\partial {\text{x}}^2}\text{for}1<\text{x}<1+\text{b}$$ (3)

Where, D_PE and D_W are the diffusivities (mm²/s) of each HOC through the polyethylene and water, respectively, x is the distance from the center of the polyethylene, l is half the thickness of the polyethylene (mm), and b is the thickness of the WBL (mm). A WBL for each passive sampler can be estimated by solving the system of equations described above using results from the PRCs. Using that information, fractional equilibrium information can be estimated for all targeted PCBs. More details about the derivatization of this model can be found in Thompson et al. (2015) and Fernandez et al. (2012).

Uncertainty in C_free can arise from either the K_PEW estimates used or from the C_PE^∞ value measured or calculated from the deployment. Cross validation experiments have shown that C_free values derived using different passive sampling mediums like silicone rubber, semi-permeable membrane devices (SPMDs), and PE often perform similarly and yield C_free estimates that are within a factor of two (Jacquet et al. 2014; Fernandez et al. 2012; Schmidt et al. 2017). These studies provide confidence in the partitioning coefficients (K_PEW) that are available for PE. In this study, K_PEWs for PCB congeners were derived from K_OW values obtained by Hawker and Connell, (Hawker and Connell 1988) using the correlation recommended by Ghosh et al. (Ghosh et al. 2014). Uncertainty in C_PE^∞ often derives from analytical error associated with C_PE^t, which is relatively small compared to that of K_PEW values. Using PRCs introduces additional measurements and estimates that contribute to the uncertainty of C_PE^∞ estimates. Often in situ, PRC-based estimates of C_PE^∞ are validated by comparing equilibrium estimated C_free values with synonymous ex situ deployments where equilibrium had been confirmed (Apell and Gschwend 2016; Gschwend et al. 2011; Fernandez et al. 2014; Schmidt et al. 2017). In general, the estimated C_free values when equilibrium had been achieved or when it had been modeled using PRC data agreed within a factor of two. These types of empirical comparisons are often performed in sediment pore water as the kinetics are more difficult to model in that system and similar experiments in surface waters would require very large grab samples and may still be inconclusive due to relatively high detection limits. Further, similar investigations with surface waters are very challenging because of contaminant depletion under ex situ conditions.

Another way to assess the performance of PRCs is to conduct a temporal study, in which pre-loaded passive samplers are exposed to contaminated water over increasing time intervals ideally until all targeted contaminants have reached equilibrium. The C_free estimated using C_PE^t and PRC results can then be compared to measured C_PE^∞ or C_PE^∞ extrapolated from the uptake curve. Recent temporal studies using PRC loaded passive samplers in quiescent sediments showed mixed results and caution the use of PRCs in field applications (Bao et al. 2016; Choi et al. 2016). To the best of our knowledge this type of study has not been conducted in the surface waters where the sampling kinetics are more easily modeled.

The primary objective of this study was to better understand (1) the accuracy of the PRC-based estimates of C_PE^∞ as a function of the fractional equilibrium attained and (2) the reproducibility of C_free estimates based on the different fractions of equilibrium attained. To evaluate the accuracy of PRC-based measurements of C_PE^∞, the temporal experimental design outlined above was performed in a series of water column deployments at three stations in the U.S. EPA Superfund site in New Bedford Harbor (MA, USA). The target contaminants were a suite of 27 PCBs contaminating the sediments of the harbor. PE passive samplers were used in the deployments along with a compliment of six ¹³C-labelled PCB PRCs. PE samplers were collected after 30, 56, 99, and 129-day deployments. This unique temporal study may yield better insight on optimal deployment times for water column passive sampling in similar environmental conditions.

Materials and Methods

Materials

Low-density PE sheeting (Ace Hardware, Oak Brook, IL, USA; 25.4 μm thick) was cut into strips (8cm × 50cm, ~ 0.5 to 3 g) and pre-cleaned in dichloromethane (DCM; 1 day), methanol (MeOH; 2 days), and stored in milli-Q water (18.3 MΩ) prior to loading PRCs. PRCs were loaded into cleaned PE in 2 L of a 4:1 MeOH:water spiking solution containing ¹³C-CB congeners: 8, 28, 52, 101, 138, and 180 for 28 days while gently mixing on an orbital shaker following Booij et al. (2002). The nominal concentration of each PRC in the spiking solution is given in SI Table S1. A piece of PE (~5 cm) was cut from each sampler to determine the PRC C_PE⁰ (i.e., concentrations in the passive sampler prior to the deployment) and stored at −20°C in the dark until further work up. PRC-loaded PE strips were then dried and woven onto stainless steel wire (Malin Co., Brookpark, OH; 20 gauge) and shaped into rings then stored at −20 °C or on ice in the dark until deployment.

All non-labeled chlorinated biphenyls (CBs): 8, 18, 28, 44, 52, 66, 70, 77, 81, 99, 101, 105, 110, 114, 118, 123, 126, 138, 153, 156, 157, 167, 169, 170, 180, 189, and 206 were purchased from Ultra Scientific (N. Kingston, RI, USA) and ¹³C labeled CBs: 8, 9, 28, 52, 101, 118, 138, 180, and 188 were purchased from Cambridge Isotope Laboratories, Inc. (Andover, MA, USA). ACS/Pesticide grade DCM and MeOH, were purchased from Honeywell (Muskegon, MI, USA) and used without further purification. Ultra-resi analyzed grade hexane (95% n-hexane) was purchased from J.T. Baker (Center Valley, PA, USA). Concentrated sulfuric acid (Fisher Scientific, Hampton, NH, USA) was used to clean-up extracts when necessary.

Field Studies

PE passive samplers were deployed in the water column in June of 2015 at two stations in New Bedford Harbor (NBH 2 and NBH4) and one station in Buzzards Bay (NBH5) (SI Figure S1). Generally, at low-energy contaminated sites, like New Bedford Harbor, the water column C_free is not expected to change substantially during deployment unless high energy events occur (e.g., storms, dredging). During this investigation, portions of New Bedford Harbor were undergoing remedial dredging (see Results and Discussion). Samplers were deployed in triplicate at each station. Four PE samplers were deployed on each mooring at each station at approximately one meter above the sediment. NBH2 is the most contaminated station followed by NBH4 with NBH5 being the least contaminated site (Pruell et al. 1990). One sampler was retrieved from each mooring at each of the four collection dates: 30, 56, 99, and 129 days. One non-PRC-loaded PE sampler and one PRC-loaded PE sampler were exposed to the atmosphere during deployment and recovery operations at all three stations to serve as a travel/field blank. Upon retrieval, PE samplers were placed in glass jars on ice during transit. Samplers were rinsed, wiped clean of visible fouling, and gently dried with laboratory tissue upon return to laboratory. Dried samplers were then stored in glass vials at −20°C in the dark until extraction.

Sample Processing and Analysis

Extraction details followed Joyce and Burgess (2018). Briefly, PE and blanks were spiked with recovery standards ¹³C-PCBs: 9, 118, and 188 (12.5 μL; 20 μg/mL) and extracted by mixing on an orbital shaker in DCM (~90mL, overnight two times). Combined extracts were concentrated using a TurboVap (Zymark, Hopkinton, MA, USA; 5–10 psi, 35°C), exchanged to hexane, concentrated down to 1 mL, and spiked with internal standard, CB198 (12.5 μL; 20 μg/mL). When particularly dark in color indicating the presence of natural organic matter, extracts were cleaned-up by acid washing with concentrated H₂SO₄ (200 μL). Extracts were stored in amber glass, crimp cap vials at −20 °C in the dark until analysis. Samples were analyzed using an Agilent 7890A digital gas chromatograph coupled to a 5975 mass selective detector (Santa Clara, CA) operating in the electron ionization (70 eV) and selected ion-monitoring mode (GC-EI/MS).

Data Analysis and Quality Control

Chromatograms were integrated using ChemStation (E.02.00.493; Santa Clara, CA, USA). PE extracts were quantified using internal standards on a seven-point calibration curve. No target contaminants were detected in control PE blanks. Twelve of the 5-cm, pre-cut pieces per thickness were extracted to determine C_PRC⁰. For each PRC, initial measurements of the PRCs in the subsamples of passive sampler (i.e., C_PRC⁰) agreed within 15%. Recoveries for ¹³C-CB9, 118, and 188 were 52±4, 87±2, and 74±4, respectively. All targeted PCB congeners were corrected for surrogate recovery according to SI Table S2. Using Equation 1, C_free was calculated using a compound specific K_PEW (SI Table S2). To simplify data interpretation, six congeners from different chlorination levels (i.e., CB8 (2 Cl); CB28 (3 Cl); CB52 (4 Cl); CB101 (5 Cl); CB138 (6 Cl); CB180 (7 Cl)), are discussed in detail below. These target congeners correspond to the six ¹³C-labelled PRCs and suggest the resulting estimates of C_free may be more accurate than for target congeners without analogous ¹³C-PRCs. Seawater temperature ranged from approximately 17˚C to 21˚C during the deployments (https://www.nodc.noaa.gov/dsdt/cwtg/natl.html visited 1 February 2020). This variation in temperatures is not expected to affect PCB partitioning to the passive sampling polymer substantially (Ghosh et al. 2014) and was not taken into account when estimating C_free values. Unless otherwise indicated, data are presented as the mean ± standard deviation for n = 3. For consistency, mean and standard deviations are also reported for station NBH2 days 99 and 129 although only two replicates were available (see below).

Results and Discussion

Behavior of Target and PRC PCBs

As anticipated, concentrations of target contaminants in the PE (C_PE^t) increased (NBH4; Figure 1a) and the PRC concentrations decreased with increasing deployment time (Figure 1b; NBH2 and NBH5 results are shown in SI Figures S2 and SI Figure S3, respectively). Remedial dredging occurred upstream from NBH 2 between the 56 and 99-day time points and resulted in the loss of one of the sampler replicates so the 99- and 129-day time points had only duplicates available for analysis. Normalizing each C_PE^t by the concentration in PE at 129 days (C_PE¹²⁹) allowed all target compound time point profiles to be compared to one another despite C_PE values ranging two orders of magnitude.

Figure 1.

Relative uptake of selected target CBs into PE at all four time points (a), and the fraction of PRC retained in PE ($C_{PRC}^t$) at all four time points (b) at station NBH4. In order that all of the data can be observed, target CB data in (a) are normalized by $C_{PE}^{129}$ while PRC data in (b) are normalized by $C_{PRC}^0$. Results for stations NBH2 and NBH5 are in the SI.

Comparing C_PE^t/C_PE¹²⁹ at NBH4 at all time points revealed some expected as well as some unexpected trends (Figure 1a). As deployment time increased, concentrations of target contaminants increased with respect to C_PE¹²⁹. This trend is especially true for CB 101, 138, and 180. Concentration ratios for CB 8, 28, and 52 were more variable. For these congeners, the concentration ratios were similar for the 30- and 56-day time points and then nearly doubled for the 99- and 129-day time points. The comparable concentrations for the 30- and 56-day time points suggest that these compounds had reached, or were close to, equilibrium. The increase between 56 and 99 day, as well as fractions of PRCs retained, suggest that equilibrium may not have been attained, or there was a change in the C_free during the deployment (possibly due to the dredging and resulting resuspension of contaminated sediments).

Similar behavioral trends were observed with the PRCs (¹³C-CB52, ¹³C-CB101, ¹³C-CB138, and ¹³C-CB180) (Figure 1b). Overall, the PRCs behaved as expected such that at each time point the fraction of PRC retained increased with molecular weight, and the fraction retained for each PRC decreased over time. However, contrary to mass-transport model predictions ¹³C-CB8 did not completely desorb by the 30-day time point. A quantifiable amount of the PRC was detected even after 129 days. The f_ret was greatest at 30 day and was reduced by about one half for the remaining time points (SI Table S3). ¹³C-CB28 also exhibited f_ret values that were higher than expected based on mass transport modeling and were often higher than those of ¹³C-CB52, although the f_ret did decrease as deployment time increased (SI Table S3). It is unclear why the PE retained these two PRCs at higher than expected fractions. While biofouling of the passive sampler’s surfaces was observed during some deployments, it is unlikely this explains the behavior of the ¹³C-CB8 and ¹³C-CB28 PRCs as the other larger PRCs behaved as expected. If biofouling was affecting the PRCs systematically, the larger PRCs would have also shown anomalous behavior which was not observed.

Estimation of Fractional Equilibrium

Sufficient data was collected to estimate f_eq using a diffusion-based mass transfer model in the water column (Thompson, Hsieh, and Luthy 2015; Tcaciuc, Apell, and Gschwend 2015; Fernandez et al. 2014). Measured f_ret ranged from 0 to 1 depending on the PRC and the time point (SI Tables S3). Model-estimated water boundary layers ranged from 65 to 190 μm in thickness and provided parameters for estimating f_eq for all target PCBs. Fractional equilibrium values from the diffusion model were then used to estimate C_free values for all targeted PCBs at each station and time point (SI Tables S4, S5). The C_free estimates ranged up to five orders of magnitude for a given target contaminant at the three different stations.

Sampling was fastest (i.e., most rapid approach to equilibrium values based on f_eq > 0.95) at NBH5 followed by NBH2 and slowest at NBH4. These rates of achieving equilibrium likely reflect the energy levels (i.e., mixing) at each station with NBH5 being the most dynamic and NBH4 the least. While the f_ret for the PRCs were used to calibrate the mass transfer model, f_eq for all PRCs were estimated along with those for target contaminants. Estimated f_eq values from the diffusion model ranged from 0.02 to 1. The model estimated CB8 had reached equilibrium in all deployed PE at all stations and time points and that both CB28 and CB52 had reached equilibrium at all stations by day 56. By day 129, the model predicted that PCBs with up to five chlorines had reached equilibrium, and analytical measurements of PRCs showed that all compounds except CB138 and CB180 (Cl₇ and Cl₈) were equilibrated.

Temporal Behavior of C_free

Variation in C_free estimates was investigated by estimating C_free from deployments of varying deployment times. The ratio of (C_free^t)⁄(C_free¹²⁹) allowed an overlay of all three stations on one plot to investigate any temporal trends over the 129-day sampling period (Figure 2). Remedial dredging occurred near station NBH2 between the 56- and 99-day sampling points. This unexpected dredging event may have contributed to variation in observed C_free concentrations in the temporal data as New Bedford Harbor is a relatively shallow estuary (Pruell et al. 1990); consequently, sediment resuspension events that can affect the water column concentrations of PCBs may occur as a result of many phenomena including storms, ship traffic and remedial and navigational dredging.

Figure 2.

Relative C_free time series using normalization by $C_{PE}^{129}$ for selected target contaminants at all three sampling stations.

The estimated C_free remained relatively constant for all four time points at all three stations for CB28, CB52 and CB101, as measurements varied less than 20%. However, CB8 exhibited a concentration fluctuation at stations NBH2 and NBH4 between the 56- and 99-day time points. Estimated C_free concentrations of CB8 were consistent for the 30- and 56-day time points and increased at the 99-day time point then decreased to concentrations similar to the 30- and 56-days by 129 days. CB8 is more water soluble than the other target PCBs and, of the target contaminants we investigated, this congener has a faster sampling rate in polyethylene (Schwarzenbach et al. 2003). These characteristics suggest C_free values for CB8 will demonstrate a higher degree of sensitivity to environmental fluctuations in concentration when compared to the other target congeners.

C_free estimates became more variable with time for target contaminants CB138 and CB180. Further, CB138 and CB180 C_free estimates were the highest and had the largest standard deviations at the 30-day time point at all three stations. The temporal patterns were similar for both of these target contaminants with the highest C_free estimates at day-30, followed by day-99, with days-56 and day-129 having similar concentrations of C_free. The triplicates from the 30-day time point had the highest variability of C_PE^t which more than doubled when C_PE^∞ was estimated. Analytical bias was investigated as a possible culprit for the high concentration and high variability for the day-30 time points. Surrogate standards at 30-days showed similar recoveries (within 20%) as the later time points and sample and internal standard area counts and intensities were also similar ruling out any significant over corrections. Intermittently injected calibration standards throughout the batch sequence showed high reproducibility ruling out significant calibration error. Closer inspection of the latter three time points indicated agreement within about 30% for CB138 and 50% for CB180. Consequently, it is believed that the high estimate of C_free and high variability in C_free is a result of errors associated with small fractional equilibrium values causing an over-extrapolation of C_free.

Variability in C_free

The C_free estimates of target PCBs were not expected to change over the course of the deployment. As discussed in the section Temporal Behavior of C_free, several different repeatable trends were observed in the temporal plots of each of the target contaminants at the different stations (Figure 2). Relative standard deviations (RSDs) were calculated for the 27 PCBs, from C_free values derived at each of the four time points (n = 12 for NBH4 and NBH5, and n = 10 for NBH2). These RSDs were then plotted against the compound’s log K_PEW to investigate trends (Figure 3). Previous reports suggest that PRC-corrected Cfree values from passive sampling with PE have an error of about 20–25% (Apell and Gschwend 2016 and Joyce and Burgess 2018). In general, the RSD is higher than 20% for the lighter PCBs and drops down for the mid-range CBs with four chlorines, then starts to increase as log K_PEW increases. This observation correlates well with what was observed in Figure 2: CB8 showed elevated C_free values at 99 day, which may have been due to active dredging that occurred during the investigation, while CB28, CB52, and CB101 show good agreement across all four timepoints, and CB138 and CB180 show elevated C_free estimates at the 30 day time point. Figure 3 illustrates the trends observed with all 27 PCBs showing that most of the C_free measurements had an RSD greater than the standard of 20–25% over the 129-day time series.

Figure 3.

Relative standard deviation (%RSD) for target contaminant C_free estimates plotted against PE partitioning coefficients (log K_PEW) for all data differentiated by station with and without the 30-day deployment included.

In Figure 3, the RSD was plotted by station (various marker shapes/colors) with and without 30-day time point data (open markers). This break down illustrates two trends: the day 30 data increased the RSD for higher chlorinated CBs and that lower concentrations, like those at NBH5, also resulted in higher RSDs. For CBs with 3 chlorines, or CBs where equilibrium was estimated to have been achieved by 30 day, the RSD for C_frees was comparable when the 30 day values were included. Results from the 30-day time point begin to diverge from the other time points for CBs with six or more chlorines. These high C_free values relative to those determined for later timepoints often were estimated using a fractional equilibrium of about 0.1 or lower which corresponds to a correction factor of about ten or more. On average, CB180 had a fractional equilibrium of 0.02, or a correction factor of 50, at the 30-day time point. Even a small error associated with these estimates could result in large differences in estimated C_free concentrations, which is evidenced by the larger error bars at the day 30 estimates for CB138 and CB180 in Figure 2. By 129 days, CB138 was predicted to have attained a fractional equilibrium of about 0.75 at NBH2, where there was a 10% difference between the two measured C_free values compared to an RSD of 29% at day 30 (f_eq was about 0.10). A similar scenario was observed for CB180 as well as the other highly chlorinated PCBs (chlorine number > 6) measured but not reported in this work.

As discussed previously, CB138 and CB180 did not reach equilibrium over the 129 days and yielded the most variable results, and the smaller f_eq values associated with these congeners at the 30-day time point had relatively large RSD values amongst the replicates. The C_free estimates at 30 day would have been equal to those of the equilibrium corrected day 129 values if CB138 had reached 17–30% equilibrium and CB180 had reached 7–11%. Table 1 summarizes the f_eq from the corresponding PRC measurement (f_eq = 1-f_ret), compared to the f_eq derived from the diffusion model, and the f_eq that would have generated the same C_free as 129-day. The diffusion model used to predict f_eq fit the measured data well but can over-predict C_free based on a small amount of error that may be associated with measurements of C_PRC⁰.

Table 1.

Observed, modeled, and estimated fractional equilibriums (f_eq) after 30 day deployment for CB101, CB138, and CB180 at three stations.

PCB Congener	Station	Observed feq	Equilibrium Diffusion Modeled feq	feq to = ${\mathit{C}}_{\mathit{P}\mathit{E}}^{129}$
CB101	NBH2	0.41±0.04	0.35±0.10	0.65
	NBH4	0.41±0.04	0.31±0.03	0.38
	NBH5	0.53±0.02	0.49±0.10	0.42
CB138	NBH2	0.15±0.03	0.12±0.04	0.22
	NBH4	0.17±0.08	0.10±0.01	0.17
	NBH5	0.26±0.02	0.16±0.05	0.31
CB180	NBH2	0	0.02±0.01	0.07
	NBH4	0	0.024±0.003	0.06
	NBH5	0.06±0.02	0.04±0.01	0.11

Passive Sampling C_free Estimates versus Measured Water Concentrations

C_free estimates were also compared to water concentrations (total and dissolved). In a separate investigation, discrete, one-liter grab water samples were independently taken and analyzed by Battelle (Norwell, MA, USA) in August 2015 to measure total and dissolved fractions of PCBs (personal communication). Samples were taken near NBH2 and in close proximity to the day 99 PE collection date. After filtration through a solvent-rinsed glass fiber filter (~1 μm pore size), total concentrations were determined by liquid-liquid extraction and dissolved fractions were assessed by solid phase extraction using a C-18 reverse-phase chromatography cartridge as described in Burgess et al. (2015). Detection limits for each congener were around 1 ng/L for both total and dissolved fractions.

Due to the relatively high detection limits associated with discrete grab sampling, a small number of congeners were detected in both studies. Twenty congeners were measured in the total concentration that were also measured in this study and eight congeners detected in the dissolved fraction were measured in this study. It is worth noting that the total and C-18 based measurements can suffer from many sampling artifacts not affecting passive sampling. These include losses of target contaminants to sampling container surfaces and over-estimations of target contaminant concentrations resulting from accidentally extracting colloids and very small sediment particles (Adams et al. 2003). Together, these artifacts cause uncertainty in these measurements that do not occur with passive sampling. Comparisons of the two measurements with the C_free estimates from the day-99 deployment are illustrated in Figure 4. As expected, the total concentrations were larger than the corresponding C_frees calculated herein. On average, the total concentrations were about forty times higher than the C_free estimates, where the ratio ranged from two to about 75 and increased with hydrophobicity (Figure 4). The measured dissolved concentrations based on C-18 isolation were, on average, within a factor of 2.5 of the estimated C_free concentrations measured in this study. More specifically, five of the eight mutually measured compounds agreed within a factor of two. The estimated C_frees from this study were larger by a factor of 18, 2.5, and 3.7 for congeners CB28, CB52, and CB110, respectively. According to our calculated C_free values, CB70, CB99 and CB118 should have also been detected in the dissolved measurements but were reported below detection limits. The larger differences in concentration for CB 28, CB52, and CB110 may be a result of errors in the partitioning coefficients or the analytical chemical measurements. However, this error likely will not completely explain the large difference in CB28 concentrations. The estimated C_free values were not increased due to PRC-correction, so over-correction did not contribute to this difference.

Figure 4.

Discrete water concentration measurements versus C_free measured over 99 days by passive sampler at or near NBH2.

It is also important to recognize that these discrete samples are representative of the moment that they were taken. The passive sampling derived C_free is a time integrated measurement dependent upon how rapidly a given target contaminant achieves equilibrium with the passive sampler. Lower molecular weight target PCBs, in this study, may reflect C_free at the end of the deployment while higher molecular weight target PCBs will reflect more of the entire deployment period, which in this specific comparison was 99 days. Many of the lighter more mobile congeners (e.g. CB8, CB18, and CB28) showed a large C_free over 99 days which decreased downward to concentrations comparable to the 30- and 56-day deployments by the day 129 sampling period. If concentrations had increased temporarily due to dredging or a weather event and then decreased by the time discrete sampling occurred, it would not be reflected in the discrete sample but would be incorporated into the passive sampling derived C_free.

Implications

Overall, this unique temporal study has shown that using PRCs yields reproducible C_free values that are within a factor of three for all estimated fractional equilibrium values generated using a diffusion-based mass transfer model. C_free values that were estimated using a fractional equilibrium less than 0.10, often resulted in C_free values that were three times larger than estimates that had been allowed longer deployment times and required smaller corrections (i.e., closer to equilibrium). This variability suggests that large corrections (i.e., low fractional equilibrium values) may yield less accurate C_PE^∞ and C_free estimates. Contaminants that had fractional equilibrium values around 0.05 (PCBs with five or more chlorines) after a 30-day deployment occasionally resulted in C_free values that were almost twice as high as estimates that required smaller corrections. Contaminants that had achieved equilibrium by 30 days or by 56 days showed good agreement over all four sampling points. Therefore, to reduce variability and generate more reliable C_free values it is recommended that when targeting very hydrophobic, slow-equilibrating contaminants, like PCBs with more than six chlorines or contaminants with a log K_OW ≥ 6.5, a deployment time of longer than 30 days may be beneficial and prudent. The consistency of estimated C_free for contaminants at equilibrium suggests that the sampler integrity will be intact over a longer sampling period.